Authors - Department of Veterans' Affairs · Web viewUS forces withdrew from Iraq by December...
-
Upload
nguyendang -
Category
Documents
-
view
215 -
download
0
Transcript of Authors - Department of Veterans' Affairs · Web viewUS forces withdrew from Iraq by December...
Systematic reviews of psychological disorders, multisymptom illness and chronic fatigue syndrome in veterans deployed to the Gulf War, Afghanistan or Iraq War
Technical Report 2015
Authors
Monash University
Dr Helen Kelsall
Professor Malcolm Sim
Professor Andrew Forbes
University of Melbourne
Professor Mark Creamer
Publications
Kelsall HL, Wijesinghe MSD, Creamer MC, McKenzie DP, Forbes AB, Page MJ, Sim MR. Alcohol
use and substance use disorders in Gulf War veterans compared to non-deployed military
personnel. Epidemiologic Reviews 2015; 37:38-54 38-54 doi:10.1093/epirev/mxu014
Blore JD, Sim MR, Forbes AB, Creamer MC, Kelsall HL. Depression in Gulf War veterans. A
systematic review and meta-analysis. Psychological Medicine 2015; 45(8):1565-1580
doi:10.1017/S0033291714001913
2
AcknowledgementsThis research was supported by Applied Research Program Grant ARP1122 from the Australian
Department of Veterans’ Affairs (DVA).
We would like to acknowledge the contribution of several researchers and other individuals who
contributed to aspects of the project:
Ms Lorena Romero, Senior Librarian at the Alfred Health Ian Potter Library, for her advice in
developing and refining the search strategy for the multiple databases.
Dr Jed Blore for undertaking the initial search under the search strategy and his major
contribution to the paper on depression in Gulf War veterans.
Dr Supun Wijesinghe for his contribution to the systematic review and meta-analysis of alcohol
use and substance use disorders in Gulf War veterans.
Ms Stella Gwini, who contributed to the review and meta-analysis of multisymptom illness.
Dr Matthew Page for his assistance in completing the statistical analyses.
3
Table of contentsAuthors...............................................................................................................................................2
Publications.....................................................................................................................2
Acknowledgements............................................................................................................................3
Table of contents................................................................................................................................4
List of Tables......................................................................................................................................6
List of Figures.....................................................................................................................................8
Abbreviations...................................................................................................................................11
1 Introduction...............................................................................................................................13
2 Aims..........................................................................................................................................15
3 Method......................................................................................................................................16
3.1 Health outcomes considered and definitions...................................................................16
3.2 Search strategy and selection criteria..............................................................................17
3.3 Study selection and data extraction.................................................................................19
3.4 Risk of bias assessment..................................................................................................21
3.5 Analytical strategy............................................................................................................21
4 Psychological disorders in Gulf War, Afghanistan and Iraq War veterans compared to non-
deployed military personnel......................................................................................................24
4.1 Literature search results..................................................................................................24
4.2 Depression in Gulf War veterans compared to non-deployed military personnel............26
4.2.1 Results.............................................................................................................26
4.2.2 Key findings.....................................................................................................28
4.3 Depression in Afghanistan/Iraq War veterans compared to non-deployed military
personnel.........................................................................................................................34
4.3.1 Results.............................................................................................................34
4.3.2 Key findings.....................................................................................................45
4.4 PTSD in Gulf War veterans compared to non-deployed military personnel....................46
4
4.4.1 Results.............................................................................................................46
4.4.2 Key findings.....................................................................................................62
4.5 PTSD in Afghanistan/Iraq War veterans compared to non-deployed military personnel.63
4.5.1 Results.............................................................................................................63
4.5.2 Key findings.....................................................................................................78
4.6 Alcohol use and substance use disorders in Gulf War and Afghanistan/Iraq War veterans
compared to non-deployed military personnel.................................................................79
4.6.1 Results.............................................................................................................79
4.6.2 Key findings.....................................................................................................96
4.7 Generalised anxiety disorder in Gulf War and Afghanistan/Iraq War veterans compared
to non-deployed military personnel..................................................................................97
4.7.1 Results.............................................................................................................97
4.7.2 Key findings...................................................................................................107
5 Multisymptom illness in Gulf War, Afghanistan and Iraq War veterans compared to non-
deployed military personnel....................................................................................................108
5.1 Literature search results................................................................................................108
5.2 Multisymptom illness in Gulf War veterans compared to non-deployed military personnel
109
5.2.1 Results...........................................................................................................109
5.2.2 Key findings...................................................................................................116
6 Chronic fatigue syndrome in Gulf War, Afghanistan and Iraq War veterans compared to non-
deployed military personnel....................................................................................................117
6.1 Literature search results................................................................................................117
6.2 CFS in Gulf War veterans compared to non-deployed military personnel.....................118
6.2.1 Results...........................................................................................................118
6.2.2 Key findings...................................................................................................119
7 Discussion...............................................................................................................................124
8 Implications of findings............................................................................................................127
9 Implications for future research...............................................................................................129
5
10 Conclusion...............................................................................................................................130
11 References..............................................................................................................................132
6
List of TablesTable 1 Characteristics of eligible studies comparing prevalence of depression in Gulf War
veterans and non-deployed military personnel................................................................................29
Table 2 Sensitivity analyses excluding each study one by one for studies of depression in
Afghanistan/Iraq War veterans compared to non-deployed military personnel...............................38
Table 3 Characteristics of eligible studies comparing prevalence of depression in Afghanistan/Iraq
War veterans and non-deployed military personnel.........................................................................40
Table 4 Sensitivity analyses excluding each study one by one for studies of PTSD in Gulf War
veterans compared to non-deployed military personnel..................................................................51
Table 5 Characteristics of eligible studies comparing prevalence of PTSD in Gulf War veterans and
non-deployed military personnel......................................................................................................53
Table 6 Sensitivity analyses excluding each study one by one for studies of PTSD in
Afghanistan/Iraq War veterans compared to non-deployed military personnel...............................68
Table 7 Characteristics of eligible studies comparing prevalence of PTSD in Afghanistan/Iraq War
veterans and non-deployed military personnel................................................................................70
Table 8 Characteristics of eligible studies comparing prevalence of alcohol/substance use
disorders in Gulf War veterans and non-deployed military personnel.............................................85
Table 9 Characteristics of eligible studies comparing prevalence of alcohol/substance use
disorders in Afghanistan/Iraq War veterans and non-deployed military personnel..........................90
Table 10 Characteristics of eligible studies comparing prevalence of generalised anxiety disorder
(GAD) in Gulf War veterans and non-deployed military personnel................................................103
Table 11 Characteristics of eligible studies comparing prevalence of generalised anxiety disorder
(GAD) in Afghanistan/Iraq War veterans and non-deployed military personnel............................105
Table 12 Sensitivity analyses excluding each study one by one for studies of multisymptom illness
in Gulf War veterans compared to non-deployed military personnel.............................................111
Table 13 Characteristics of eligible studies comparing prevalence of multisymptom illness in Gulf
War veterans and non-deployed military personnel.......................................................................112
7
Table 14 Sensitivity analyses excluding each study one by one for studies of CFS in Gulf War
veterans compared to non-deployed military personnel................................................................119
Table 15 Characteristics of eligible studies comparing prevalence of chronic fatigue syndrome
(CFS) and CFS-like illness in Gulf War veterans and non-deployed military personnel................120
Table 16 Main meta-analysis summary odds ratios for psychological disorders, multisymptom
illness and chronic fatigue syndrome (CFS) in Gulf War, Afghanistan/Iraq War veterans compared
with non-deployed personnel.........................................................................................................124
8
List of FiguresFigure 1: PRISMA flowchart of the systematic review of psychological disorders in Gulf War
veterans, Afghanistan/Iraq War veterans, 1990-2014......................................................................25
Figure 2 Random effects meta-analysis of depression in Gulf War veterans compared to non-
deployed military personnel.............................................................................................................27
Figure 3 Random effects meta-analysis of dysthymia or chronic dysphoria in Gulf War veterans
compared to non-deployed military personnel.................................................................................28
Figure 4 Random effects meta-analysis of depression in Afghanistan/Iraq War veterans compared
to non-deployed military...................................................................................................................34
Figure 5 Random-effects meta-analysis of depression in Afghanistan/Iraq War veterans compared
to non-deployed military personnel, subgrouped by type of outcome measure...............................35
Figure 6 Random-effects meta-analysis of depression in Afghanistan/Iraq War veterans compared
to non-deployed military personnel, subgrouped by type of service................................................36
Figure 7 Random-effects meta-analysis of depression in Afghanistan/Iraq War veterans compared
to non-deployed military personnel, subgrouped by risk of bias......................................................37
Figure 8 Random-effects meta-analysis of depression in Afghanistan/Iraq War veterans compared
to non-deployed military personnel, subgrouped by adjustment of odds ratios...............................38
Figure 9 Funnel plot for the random-effects meta-analysis of depression in Afghanistan/Iraq War
veterans compared to non-deployed military personnel..................................................................39
Figure 10 Random-effects meta-analysis of PTSD in Gulf War veterans and non-deployed military
personnel.........................................................................................................................................46
Figure 11 Random-effects meta-analysis of PTSD in Gulf War veterans and non-deployed military
personnel, subgrouped by type of outcome measure......................................................................47
Figure 12 Random-effects meta-analysis of PTSD in Gulf War veterans and non-deployed military
personnel, subgrouped by type of service.......................................................................................48
Figure 13 Random-effects meta-analysis of PTSD in Gulf War veterans and non-deployed military
personnel, subgrouped by risk of bias.............................................................................................49
9
Figure 14 Random-effects meta-analysis of PTSD in Gulf War veterans and non-deployed military
personnel, subgrouped by adjustment of odds ratios......................................................................50
Figure 15 Funnel plot for the random-effects meta-analysis of PTSD in Gulf War veterans and non-
deployed military personnel.............................................................................................................52
Figure 16 Random-effects meta-analysis of PTSD in Afghanistan/Iraq War veterans and non-
deployed military personnel.............................................................................................................63
Figure 17 Random-effects meta-analysis of PTSD in Afghanistan/Iraq War veterans and non-
deployed military personnel, subgrouped by type of outcome measure..........................................64
Figure 18 Random-effects meta-analysis of PTSD in Afghanistan/Iraq War veterans and non-
deployed military personnel, subgrouped by type of service...........................................................65
Figure 19 Random-effects meta-analysis of PTSD in Afghanistan/Iraq War veterans and non-
deployed military personnel, subgrouped by risk of bias.................................................................66
Figure 20 Random-effects meta-analysis of PTSD in Afghanistan/Iraq War veterans and non-
deployed military personnel, subgrouped by adjustment of odds ratios..........................................67
Figure 21 Funnel plot for the random-effects meta-analysis of PTSD in Afghanistan/Iraq War
veterans and non-deployed military personnel................................................................................68
Figure 22 Random-effects meta-analysis illustrating log-transformed odds ratios of alcohol use
disorders in Gulf War veterans and non-deployed military personnel.............................................80
Figure 23 Random-effects meta-analysis illustrating log-transformed odds ratios of any substance
use disorders in Gulf War veterans and non-deployed military personnel.......................................82
Figure 24 Random-effects meta-analysis illustrating log-transformed odds ratios of alcohol use
disorders in Afghanistan/Iraq War veterans and non-deployed military personnel..........................83
Figure 25 Random-effects meta-analysis illustrating log-transformed odds ratios of substance use
disorders in Afghanistan/Iraq War veterans and non-deployed military personnel..........................84
Figure 26 Random-effects meta-analysis illustrating log-transformed odds ratios of generalised
anxiety disorder in Gulf War veterans and non-deployed military personnel...................................98
Figure 27 Random-effects meta-analysis of generalised anxiety disorder in Gulf War veterans and
non-deployed military personnel, subgrouped by case definition....................................................99
10
Figure 28 Random-effects meta-analysis of generalised anxiety disorder in Gulf War veterans and
non-deployed military personnel, subgrouped by risk of bias........................................................100
Figure 29 Random-effects meta-analysis illustrating log-transformed odds ratios of generalised
anxiety disorder in Afghanistan/Iraq War veterans and non-deployed military personnel.............101
Figure 30 Random-effects meta-analysis of GAD in Afghanistan/Iraq War veterans compared to
non-deployed military personnel, sensitivity analysis excluding Bray (2006)................................102
Figure 31 Flowchart of the systematic review of multisymptom illness (MSI) Gulf War veterans,
Afghanistan/Iraq War veterans, 1990-2014...................................................................................108
Figure 32 Random-effects meta-analysis of multisymptom illness in Gulf War veterans and non-
deployed military personnel...........................................................................................................110
Figure 33 Funnel plot for the random-effects meta-analysis of multisymptom illness Gulf War
veterans and non-deployed military personnel..............................................................................111
Figure 34 PRISMA flow diagram for chronic fatigue syndrome (CFS) in Gulf War, Afghanistan, Iraq
War veterans..................................................................................................................................117
Figure 35 Random-effects meta-analysis illustrating log-transformed odds ratios of chronic fatigue
syndrome in Gulf War veterans and non-deployed military personnel..........................................118
11
AbbreviationsAUDIT World Health Organization Alcohol use disorder identification testARP (Department of Veterans’ Affairs) Applied Research ProgramAUD Alcohol use disordersBDI Beck Depression IndexBMI Body Mass IndexCAGE CAGE QuestionnaireCAPS Clinician Administered PTSD ScaleCES-D Center for Epidemiologic Studies Depression ScaleCFS Chronic fatigue syndromeCI Confidence intervalCIDI World Health Organization Composite International Diagnostic InterviewCIDI–SF World Health Organization Composite International Diagnostic Interview
Short FormCDC Centers for Disease Control and Prevention, Atlanta, Georgia(combined) any SUD (combined) any substance use disorder refers to alcohol use disorders
and other substance use disorders combinedDoD (US) Department of DefenseDSM Diagnostic and Statistical Manual of Mental Disorders, American
Psychiatric AssociationFE MA Fixed effects meta-analysis (as appears in the funnel plot figures)GAD Generalised anxiety disorderGW Gulf WarGWV/s Gulf War veteran/sICD-9 International Classification of Diseases, 9th RevisionIOM Institute of MedicineMDD Major depressive disorderMeSH Medical Subject HeadingsMEAO Middle East Area of OperationsM-PTSD Mississippi scale for combat-related PTSDMSI Multisymptom illnessNCO Non-commissioned officerNG National GuardNG/Res National Guard/ReserveOEF Operation Enduring Freedom (Afghanistan)OIF Operation Iraqi FreedomOR Odds ratioPCL PTSD Checklist
12
PCL-C PTSD Checklist – Civilian versionPCL-M PTSD Checklist – Military versionPDHA Post-Deployment Health AssessmentPHQ Patient Health QuestionnairePRIME-MD PHQ Primary Care Evaluation of Mental Disorders Patient Health QuestionnairePTSD Posttraumatic stress disorderR Regular militaryRR Relative RiskSCID Structured Clinical Interview for DSM DisordersSD Standard deviationSUD Substance use disorders (for brevity in this report is used to refer to
substance use disorders other than alcohol use disorders, e.g. opioids, sedatives, anxiolytics, cocaine, cannabis)
13
1 IntroductionIn the 1990-1991 Gulf War, a large multinational force was deployed to the Gulf area in response
to the invasion of Kuwait by Iraq on 2 August 1990. In October 2001, in response to the
September 11 attacks on the US, the United States of America (US) supported by the United
Kingdom (UK), Australia, Canada, France and Germany and other nations began an invasion of
Afghanistan (Operation Enduring Freedom). In March 2003, the Iraq War began (Operation Iraqi
Freedom) when a combined force from the US, UK, Australia and Poland invaded Iraq. US forces
withdrew from Iraq by December 2011.
The forces deployed by different countries differed. For example, in contrast to US service
members and UK personnel deployed during the 1990-1991 Gulf War who were predominantly
army land based forces, Australian personnel were primarily involved in sea and air surveillance
(around 85% of deployed personnel were in the Royal Australian Navy) (1). In the Afghanistan and
Iraq War operations, the Australian deployment involved the services of Air Force, Navy, Army,
and Special Operations Task Group (2). The Army comprised the majority of deployed UK (3) and
US active-duty (4) personnel.
Psychological and physical health is important to all military personnel, deployed personnel,
veterans and their families. Studies of the relationship between deployment to the Gulf War and to
the Afghanistan/Iraq War and health outcomes suggest that veterans of these wars are at
increased risk of psychological illnesses compared to personnel not deployed to these conflicts (5-
10). Posttraumatic stress disorder (PTSD) has tended to receive greater attention and publicity
than other psychological disorders such as affective disorders, anxiety disorders such as
generalised anxiety disorder (GAD), or substance use disorders (11), yet alcohol and other
substance use disorders have also long been associated with combat experience in other theatres
of war (12-14).
There have been several previous reviews of health outcomes, but these have mainly been in Gulf
War veterans, and since these earlier reviews further studies have been conducted and so they do
not include more recently published findings, or were narrative in style rather than systematic.
Stimpson et al. (11) conducted a systematic review of psychiatric disorders in veterans of the
1990-1991 Gulf War which was published in 2003. The Stimpson review found that, despite
heterogeneity between the studies examined, the prevalence of PTSD and common mental
disorder (depression or anxiety diagnosed using a standardised assessment or self-reported
depression symptoms on a checklist) was increased in Gulf War veterans compared with a non-
14
Gulf comparison group. The search strategy for Stimpson et al.’s (11) review ended in May 2001,
with the last reviewed study being published in 2001.
Thomas et al. (2006) (15) conducted a systematic review of symptom-based conditions (termed
multi-symptom conditions in the review) in Gulf War veterans, including the Centers for Disease
Control and Prevention (CDC)-defined multisymptom illness for papers published between 1990
and 2004, and found that Gulf War deployment was strongly associated with chronic fatigue
syndrome (CFS) and Gulf War veterans were more likely to report multiple chemical sensitivity or
chronic multisymptom illness, as defined by CDC. The review showed that there was considerable
variation in the methodological quality of the studies identified, and noted that the later and larger
studies were of a higher methodological quality (15). The search strategy for the Thomas et al (15)
review concluded in May 2004 and the last reviewed study was published in 2004.
There has been further research conducted among Gulf War veterans since these review papers
were published and this justified further reviews. Numerous studies have been published on the
psychological health of Gulf War veterans, including that of Australian veterans (1, 16), and of
reports of increased rates of anxiety (17) and depression (18) in US Gulf War veterans, and
several on multisymptom illness and CFS, including those of Australian and US Gulf War veterans
(16, 19-21). An Institute of Medicine (IOM) Gulf War and Health report published in 2010 (22)
provided a comprehensive summary report of primary and secondary studies, according to its
criteria, published since the IOM 2006 report (23) on Gulf War veteran health (11). The IOM report
was a narrative review and did not include any meta-analyses.
Furthermore, there have been an increasing number of published studies reporting the
psychological health of the Afghanistan and Iraq War veteran cohorts, which are increasingly
relevant and have also reported adverse psychological health outcomes.
One review, conducted by Kok et al. (24) and published in 2012, examined PTSD in military
personnel deployed to the Afghanistan and Iraq Wars. However, the authors did not restrict
studies for inclusion based on the presence of an appropriate comparison group and therefore
could not determine the odds of deployed personnel developing PTSD compared to non-deployed
personnel. Another review, conducted by Sundin et al. (25) and published in 2010, investigated
PTSD in Iraq deployed personnel only. Like the study by Kok et al. (24), this study did not report
the odds of developing PTSD in those deploying to Iraq compared to non-deployed personnel. A
third review, conducted by Gadermann et al. (26) and published in 2012, investigated major
depression in US personnel only, and as in the previous reviews, no restrictions were placed on
studies not including a comparison group. Results were also not reported by deployment.
15
It has been widely acknowledged that exposure to combat and other wartime experiences can
have both short-term and long-term psychological and physical effects. However, reported
psychological consequences have been quite varied. Therefore, undertaking a systematic review
assists in drawing conclusions about consistency of the results of studies in relation to health
outcomes in Gulf War veterans, Afghanistan and Iraq War veterans compared to personnel who
were not deployed to the corresponding warzone or who were deployed elsewhere. Conducting
meta-analyses and presenting the output produces visual and comparable summary effect
estimates of the health outcomes under consideration in Gulf War veterans, Afghanistan and Iraq
War veterans compared with non-deployed military personnel, and quantifies these in overall
summary measures.
A pattern of increased reporting of symptoms across several body systems has been reported in
several Gulf War veteran health epidemiological studies. In 1998 Fukuda et al. (27) developed a
case definition for multisymptom illness in Gulf War veterans, also known as the CDC definition.
This has been an accepted definition adopted by many research groups to define multisymptom
illness in further Gulf War veteran health epidemiological studies. Other definitions of
multisymptom illness have been suggested by Haley et al. (1997) (28), Steele et al. (2000) (29),
Bourdette (2001) (30), and Kang et al. (2009) (31), and these either exert more stringent rules on
the classification of the illness or use variations or slightly different methods to elucidate the
patterns of multisymptom reporting.
The body of research on psychological health and, to a lesser extent on physical health, of
veterans of deployments to the 1991 Gulf War, Afghanistan and Iraq War has grown considerably
in the past few years. Publication of further research since review papers were published has
justified further reviews in relation to psychological health outcomes. Furthermore, to our
knowledge a systematic review of any literature reporting symptom based conditions of
multisymptom illness and CFS in veterans of deployments to Afghanistan and the Iraq War has not
been undertaken. A systematic review that considers all the research done on the topic in these
veteran cohorts and presents summary estimates can provide a useful summary for practitioners
and policy makers.
The purpose of this project was to conduct systematic reviews and meta-analyses of studies that
have compared psychological disorders, multisymptom illness and CFS in veterans of deployments
to the 1991 Gulf War, Afghanistan or the Iraq War, compared to military personnel who were not
deployed to these conflicts.
16
2 AimsThe overall aim of this project was to conduct systematic reviews and meta-analyses of the
international literature to compare psychological disorders (depression, PTSD, GAD, alcohol use
disorders, other substance use disorders (e.g. opioids, sedatives, anxiolytics, cocaine, cannabis)
(herein termed substance use disorders for brevity), multisymptom illness and CFS in veterans of
the Gulf War, Afghanistan and Iraq War compared with non-deployed comparison groups of the
corresponding conflicts. A further aim in the meta-analyses was to assess sources of variability, by
subgroup analyses relevant to the particular meta-analysis being undertaken, to explore the factors
most likely to result in study heterogeneity.
3 MethodThe searches, study selection and data extraction were conducted separately for psychological
disorders, multisymptom illness and CFS; but as the methods were similar for the three broad
outcome categories in Gulf War veterans, Afghanistan and Iraq War veterans, this section
describes the methods overall. The results are then presented separately for each of these three
broad outcomes.
3.1 Health outcomes considered and definitions
The health outcomes considered were:
Psychological disorders of depression (major depression and dysthymia), PTSD, GAD,
alcohol use disorders, substance use disorders and (combined) any substance use
disorder (i.e. alcohol and/or substance use disorders),
Multisymptom illness, and
CFS.
The psychological disorders reviewed are the more commonly reported psychological conditions,
and for which assessment measures were considered likely to be more homogeneous between
studies of veterans from different countries. Although GAD was not proposed as a psychological
disorder for inclusion in the Detailed Project Plan, it is the most commonly diagnosed anxiety
disorder world-wide (32) but has received relatively little attention in military or veteran populations.
There were sufficient papers to review and to conduct a meta-analysis, and thus this outcome was
included.
17
During the conduct of this project and these systematic reviews and meta-analyses, the prevailing
classification system for psychological disorders was DSM-IV (American Psychiatric Association
Diagnostic and Statistical Manual of Mental Disorders, 4th edition) (33), and both posttraumatic
stress disorder and generalised anxiety disorder were classified under Anxiety Disorders. In DSM-
V, published in 2013, posttraumatic stress disorder was classified under Trauma- and Stressor-
Related Disorders (34). Materially this does not make any difference to the findings of these
systematic reviews.
Alcohol use disorders, and substance use disorders (i.e. relating to drugs of abuse other than
alcohol such as amphetamine or similarly acting sympathomimetics, cannabis, cocaine,
hallucinogens and opioids; phencyclidine (PCP), or similarly acting arylcylcohexylamines; and
sedatives, hypnotics, or anxiolytics, as defined in the corresponding studies; herein in this report
referred to as substance use disorders) were reported separately in the systematic reviews and
meta-analyses where possible. Some studies reported an outcome that combined alcohol use
disorders and substance use disorders; and were referred to as (combined) any substance use
disorder to differentiate the outcome from substance use disorders. In a meta-analysis of any
substance use disorder, we also included studies that had reported alcohol use and substance use
disorders outcomes.
The definition used for inclusion of studies of multisymptom illness was based on an internationally
accepted definition, specifically the CDC definition (20, 27) developed by Fukuda et al. (27) and its
variants. The CDC definition identifies multisymptom illness as present if the person has one or
more chronic symptoms (for at least six months) from at least two of three categories: fatigue,
mood-cognition (symptoms of feeling depressed, difficulty remembering or concentrating, feeling
moody, feeling anxious, trouble finding the right words or difficulty sleeping) and musculoskeletal
(symptoms of joint pain, joint stiffness or muscle pain), where the latter were the two categories
identified in their factor analysis of symptoms reported in their study of US Gulf War veterans. For
this systematic review, we used the CDC definition of multisymptom illness because it represented
an internationally accepted definition of multisystem symptom reporting and was commonly
available during the period of the review.
The definition used for inclusion of studies of CFS was based on an internationally accepted
definition, specifically the 1994 international definition (19, 35) for defining CFS and other fatiguing
illnesses.
18
3.2 Search strategy and selection criteria
The scientific literature was searched for published and unpublished citations from 1 January 1990
to 24 January 2014 using multiple electronic databases MEDLINE, MEDLINE In-Process,
PsycINFO, Embase, Published International Literature on Traumatic Stress (PILOTS) and
Cochrane Reviews. In the multisymptom illness search, the System for Information on Grey
Literature in Europe (SIGLE) was also included. Additional sources searched for non-peer review
literature included the US, UK and Australian departments of veterans’ affairs and departments of
defence websites.
Studies which matched the following inclusion criteria were included:
[1] The study population consisted of military personnel deployed to the Gulf War (1990-1991),
Afghanistan (2001- ) or Iraq War (2003-2011) encompassing Navy, Army, Air Force, Marines,
Coast Guard, medical personnel, and Reservists/National Guard;
[2] Studies were published in English;
[3] The outcome of interest was depression (major depression and dysthymia), PTSD, GAD,
alcohol use disorders, substance use disorders, (combined) any substance use disorder,
multisymptom illness defined using a definition similar to the CDC definition or its variants; or
CFS;
[4] The study included a military comparison group that differed in its level of deployment
exposure to the corresponding conflict (non-deployed personnel were defined as personnel
who did not serve in the primary area of conflict as per previous systematic reviews (36), and
other conflict/other deployed personnel were defined as personnel deployed outside the
primary area of conflict or to other conflicts, e.g. Germany, Bosnia); and
[5] For inclusion in the meta-analysis, the study needed to provide enough information to
generate an odds ratio (OR) by deployment.
Studies of the following type were excluded from the review:
[1] The conflict deployed sample was of non-military personnel;
[2] Studies that were published in a language other than English;
19
[3] The sample was based on clinical or injured or treatment/help-seeking population/s
(including studies based on data from Veterans Affairs (VA) treatment facilities);
[4] Studies with no eligible military comparison group (e.g. civilians were used as a comparison
group) or where it was not clear where the population being studied had been deployed; and
[5] In the multisymptom review, studies where the definition of multisymptom condition did not
accord with a multisystem condition in accordance with the CDC criteria, e.g. a total symptom
count was used.
We developed a list of free text and Medical Subject Headings (MeSH) terms based on the three
components in the research question. The components were “A1: Gulf War, Afghanistan War and
Iraq War”, “A2: Military personnel, military veterans, military medicine, and veterans’ health” and
“B: Psychological disorders”. The final search strategy was: (A1 or A2) AND (B). Key words were
customised to each individual database. A modified portion of the search string that was used for
Medline database in searching for psychological disorders is given below.
(exp Gulf War/ OR Persian Gulf War OR Gulf* OR Desert Storm OR Desert Shield OR exp Afghan
Campaign 2001-/ OR Afghan* OR Enduring Freedom* OR OEF OR exp Iraq War, 2003-/ OR I* OR
Iraqi Freedom* OR OIF OR TELIC OR New Dawn OR OND) OR (exp Military Personnel/ OR
Military Veterans OR military* OR service personnel OR soldier* OR active duty OR deployed*)
AND (alcohol* OR substance use OR substance abuse OR substance dependence OR drug
abuse OR drug use OR drug misuse OR drug dependence OR illicit drug* OR illicit substance* OR
substance misuse OR recreational drug* OR illegal substance* OR illegal drug* OR drug disorder*
OR (substance adj2 disorder)) OR (exp depression/ OR (depress* OR dysthymi* or melancholi*))
OR (exp generalized anxiety disorder/ OR (generalised anxiety disorder OR GAD*)).
Similarly, the multisymptom illness search terms were based on the free-text and MeSH of the
condition of interest (multisymptom or multi-symptom or multiple-symptom or multiple symptom or
CMI or Gulf War Illness* or unexplained illness* or CDC or Persian Gulf syndrome* or Gulf War
syndrome* or Persian Gulf Illness* or Gulf War Illness*) and the area of operation (as defined
above).
The CFS literature search terms were based on the free-text and MeSH of the condition of interest
(chronic fatigue or chronic-fatigue or CFS or CDC-fatigue or CDC fatigue or Chalder fatigue or
Chalder-fatigue or fatigue) and the area of operation (as defined above).
20
3.3 Study selection and data extraction
The search strategy and selection of studies was conducted according to PRISMA (Preferred
Reporting Items for Systematic Reviews and Meta-Analyses) recommendations (37). Titles and
abstracts (records) from each database were entered into the reference manager software,
EndNote version X4. Duplicates were removed. These records were screened to identify studies
for full-text review by the specified inclusion and exclusion criteria above. A member of the
research team reviewed all titles and abstracts, and HK conducted a blind review of approximately
ten per cent of titles and abstracts to refine and discuss any amendments to the reviewing strategy,
and all abstracts selected for full-text review and all eligible articles. Any discrepancies were
resolved through collaboration.
Quantitative and other relevant data for each individual study were extracted by standard data
extraction forms developed for the review (descriptive data, summary measures of effect size,
precision and assessment of risk of bias) and tabulated including first author, study design, date
and location, method of data collection, sample size including female percentage, psychological
health outcome measure and case definition used in the study, participation rates, assessment of
risk of bias, overall risk of bias assessment (see later) of each individual study, outcome of interest
as a prevalence in the Gulf War, Afghanistan, Iraq War veterans and comparison group,
unadjusted association estimates (odds ratio, OR) and adjusted (for potential confounding factors
such as age, service branch, rank) association estimates (OR) of the same.
We used the following pre-defined strategy for extracting data to be included in the meta-analysis.
Studies that were derivative studies of earlier studies and did not present new data were not
included in the review. If two papers reported an overlapping sample of participants, we prioritised
inclusion of results of the larger sample in the meta-analysis. Where more than one paper from the
same study population, or the same paper, reported the same or a similar outcome measure,
priority was given to the most valid and reliable psychological outcome case definition. The priority
order was as follows: 1) structured diagnostic interview or diagnosis made by a clinician (38); 2)
screening tool; (e.g., Patient Health Questionnaire (PHQ; Beck Depression Inventory (BDI),
Alcohol use disorder identification test (AUDIT); 3) self-reported physician diagnosis (39, 40).
Data were extracted by a member of the team and checked by HK independently and any
discrepancies were resolved through discussion. Where results were given for both non-deployed
and other-deployed comparison groups, results for the non-deployed comparison group were given
priority. Reported adjusted ORs were prioritised over unadjusted ORs and unadjusted ORs over
prevalences. Where studies did not provide an OR, these were calculated using the reported
prevalence, or numbers of deployed and non-deployed veterans with and without the psychological
disorder, and subsequently entered into the meta-analysis along with the adjusted ORs from other
21
studies. When data were not presented in the desired categories of deployed vs non-deployed,
results were recalculated using the reported prevalences and numbers (n’s).
By definition in this report, Non-deployed / non-GWV refers to a military comparison group who
were not deployed to the Gulf War during the period of operations, and Non-deployed / non-
Afghanistan/Iraq War veteran refers to a military comparison group who were not deployed to the
Afghanistan/Iraq War during the period of operations.
3.4 Risk of bias assessment
The PRISMA statement (37) notes that the reporting of assessment of risk of bias in included
studies is important in the conduct of systematic reviews. ‘Risk of bias’ refers to ‘systematic error
or deviation from the truth, in results or inferences’ (41).
Overall risk of bias of the studies included in this systematic review was assessed using an
instrument developed by Hoy et al. (42) which was used for the assessment of prevalence studies
in the Global Burden of Disease Study 2010 (43).
This instrument is reported to have a high overall inter-rater agreement of 91%, with a kappa
statistic (44) representing chance-corrected agreement of 0.82 (95% confidence interval (CI): 0.76,
0.86). This tool enables an overall risk of study bias based on assessment of the risk of bias of 10
individual items (items 1-4 and items 5-10 of this instrument are based on external and internal
validity of the included study, respectively). We included an additional item on availability of, and
adjustment for, possible confounding factors since it was expected there would be considerable
variation in the extent to which the individual studies attempted to adjust for confounding factors
(11). Individual items were assessed as high and low risk of bias, and used to assess overall risk
of bias. The authors of the tool had deliberately excluded a moderate risk of bias category as
testing indicated this was being used to avoid deciding between high and low risk of bias;
subsequently inter-rater agreement improved substantially (42).
3.5 Analytical strategy
In all meta-analyses (37) we prioritised inclusion of adjusted (for possible confounders such as
age, education, branch of service, and duty status, etc) OR over unadjusted OR and unadjusted
OR over simple prevalence figures. Where data were not presented in the desired categories of
deployed v non-deployed, results were recalculated using the reported prevalences and n’s. In
cases where the prevalence of CFS was zero in one group, the OR was calculated using RevMan
software, which adds 0.5 to all cells where zero cells cause problems with computation of effect
22
estimates or standard errors (45). All the measures of association were converted into OR and
corresponding 95% CI for studies that did not present them in the original manuscripts. As
heterogeneity of outcome was expected between studies, random effects meta-analyses (46, 47),
for the outcomes under consideration were conducted to aggregate the OR across the selected
studies and produce pooled odds ratios with 95% CI based on the standard normal distribution.
After study selection for multisymptom illness, it was identified that one study had reported results
separately for males and females and hence these results were aggregated using formulas
adapted from McNeil (48).
Statistical heterogeneity was reported using the I2 statistic which indicates variability in results
across studies that is due to heterogeneity rather than chance. I2 ranges between 0% to 100% with
higher values representing greater heterogeneity (49). Corresponding P-values indicate the
significance of the heterogeneity.
Sources of variability were assessed by subgroup analyses relevant to the particular meta-
analyses being undertaken, to explore the factors most likely to result in study heterogeneity.
These included the outcome measure used (diagnostic interview; screening tool; self-reported
physician diagnosis), type of multisymptom definition, adjusted vs unadjusted odds ratios, risk of
bias (low, high), theatre of deployment (Gulf War versus Afghanistan/Iraq War), and duty status
(regular versus reservist) in Gulf War and Afghanistan/Iraq War veterans. If results were originally
presented separately for personnel deployed to Afghanistan or Iraq, they are tabulated as such.
In the multisymptom illness review and meta-analysis, two studies were identified which had been
conducted by the same research group (21, 31) and possibly with an overlap of participants.
Hence we repeated the analyses using Bayesian random effects models with `delta splitting’ to
accommodate the overlap of participants, computed using the metahdep function in the R software
(50, 51). Since results differed negligibly from the conventional random effects analysis we report
only the conventional random effects results here.
To investigate whether there was a relationship between sources of variability and the magnitude
of the odds ratios, we undertook random effects meta-regression (52). Sensitivity analyses,
excluding studies one by one for the relevant health outcome under consideration, were conducted
for both Gulf War and Afghanistan/Iraq War veteran studies.
Publication bias was assessed by generated funnel plots (53) and by conducting the Egger test for
funnel plot asymmetry (49, 54, 55), using Stata version 11.0 or 12.0 software (StataCorp LP,
College Station, Texas) (56). A P-value of <0.05 was considered statistically significant for the
Egger test indicating funnel plot asymmetry.
23
A funnel plot is a graphical display of a measure of study precision plotted, most commonly on the
vertical axis, against effect estimate, on the horizontal axis, that can be used to investigate whether
there is a link between study size and effect estimate. One possible cause of an observed
association is reporting bias (57).
The term ‘funnel plot’ arises from the fact that precision of the effect estimate increases as the size
of the study increases, therefore effect estimates from small studies will scatter more widely at the
bottom of the graph and the spread will narrow among larger studies (41). In the absence of
publication bias, i.e. smaller studies without statistically significant effects were also published; the
plot should approximately resemble a symmetrical, inverted funnel. The more pronounced the
asymmetry, the more likely it is that the amount of bias could be greater (41).
Some problems with interpretation of funnel plots have been identified (41); for example that
publication bias need not lead to asymmetry in funnel plots (41), and that interpretation should not
be based on visual inspection alone (55), and spurious asymmetry as some effect estimates are
naturally correlated with their standard errors (41). Inferences on the presence of bias or
heterogeneity need to consider different causes of funnel plot asymmetry and be informed by
contextual factors, including publication bias as a plausible explanation for the asymmetry (55). In
the systematic reviews and meta-analyses in this report, publication bias was assessed by
generated funnel plots (53), by conducting the statistical Egger test for funnel plot asymmetry (49,
54, 55), and by considered interpretation.
A forest plot is a graphical representation of the individual results of each study included in a meta-
analysis together with the combined summary meta-analysis result (57). The individual study
results are shown as squares centred on each study’s point estimate, and the horizontal line
running through each square shows the 95% CI. The subgroup analyses estimates or the overall
summary estimate from the meta-analysis and their 95% CI are shown at the bottom of each
subgroup analysis or at the bottom of the figure, represented as diamonds. The centre of the
diamond represents the pooled point estimate, and its horizontal tips represent the CI. A forest plot
also allows readers to visually see the heterogeneity among the results of the studies (57).
All meta-analyses were performed using MetaXL 1.1 for analysis of depression in Gulf War
veterans or MetaXL version 1.4 (http://www.epigear.com) (58). Meta-regressions were conducted
using the metareg command in Stata (59).
24
4 Psychological disorders in Gulf War, Afghanistan and Iraq War veterans compared to non-deployed military personnel
4.1 Literature search results
Figure 1 shows that the search for psychological health outcomes under study yielded 23,533
records, with 14,771 records remaining after removal of duplicates. Following the removal of
duplicates, titles and abstracts were screened to identify studies for full- text review by the
specified inclusion and exclusion criteria. After abstract review, 253 full-text articles were identified
for further review and were reviewed separately in relation to Gulf War or Iraq/Afghanistan
psychological disorders’ eligibility. Of these, in relation to Gulf War eligibility, 228 were excluded
and, in relation to Iraq/Afghanistan eligibility, 224 were excluded. The reasons for exclusion are
identified in Figure 1. Forty-nine eligible articles were identified reporting psychological disorders,
25 in Gulf War veterans and 24 in Afghanistan/Iraq War veterans.
25
14,771 Records After Duplicates Removed
14,771 Records Screened 13,903 Titles Excluded
868 Abstracts Assessed For Eligibility 615 Abstracts Excluded
253 Full-Text Articles Assessed For Eligibility
228 Full-Text Gulf War and 229 Full-Text Afghanistan/Iraq Articles Excluded
35 Gulf War and 77 Afghanistan/Iraq articles had ineligible comparison group or case definition
156 non-Gulf study or Gulf derivative study reported elsewhere
104 non-Afghanistan/Iraq study or Afghanistan/Iraq derivative study reported elsewhere
19 Gulf War and 20 Afghanistan/Iraq treatment seeking sample
18 Gulf War and 28 Afghanistan/Iraq not original research or OR not calculable
49 Eligible Articles Reporting Psychological Disorders
25 Gulf War veterans
24 Afghanistan/Iraq War veterans
Depression
14 Gulf War
13 Afghanistan/Iraq (10 in meta-analysis)
23,533 Records Identified Through Database Search
Alcohol/Substance use
9 Gulf War
9 Afghanistan/Iraq (7 in meta-analysis)
Generalised anxiety disorder
5 Gulf War
3 Afghanistan/Iraq
PTSD
20 Gulf War (18 in meta-analysis)
18 Afghanistan/Iraq (16 in meta-analysis)
Figure 1: PRISMA flowchart of the systematic review of psychological disorders in Gulf War veterans, Afghanistan/Iraq War veterans, 1990-2014
27
4.2 Depression in Gulf War veterans compared to non-deployed military personnel
4.2.1 Results
The application of the inclusion and exclusion criteria yielded 201 abstracts for further review
(Figure 1). Of these, 34 were excluded based on the abstract, leaving 167 full-text titles for review.
Twenty-five articles reported on psychological disorders in Gulf War veterans; of these, 14 (1, 6-8,
29, 31, 60-67) reported depression as an outcome and met criteria for inclusion in the quantitative
synthesis. Of these 14 studies, four (1, 6, 66, 67) used structured diagnostic interviews to
determine caseness of major depressive disorder (MDD), dysthymia, or both (three of four used
the Composite International Diagnostic Interview (CIDI) with DSM-IV criteria whilst Wolfe et al.
used the Structured Clinical Interview for DSM Disorders (SCID) with DSM-IIIR criteria); seven (7,
8, 31, 60-62, 65) used depression screening tools and three (29, 63, 64) used self-reported
physician diagnosis.
Six of the 14 studies did not present adjusted ORs (6, 8, 31, 60, 61, 65). We attempted contact
with the authors to request this information; however we were unable to obtain adjusted ORs for
any of the 6 studies. Table 1 summarises these 14 studies in the order of the case definition
hierarchy given previously and within each grouping the studies were ordered by year of
publication. The same order was followed in the forest plots (Figure 2 and Figure 3).
Eight of the 14 studies received a high overall risk of bias assessment (6, 29, 60-65). None of the
four studies using structured diagnostic interviews were assessed as having a high overall risk of
bias.
Depression and major depression meta-analysis
The forest plot of the studies reporting depression (Figure 2) indicates an increased overall odds
for Gulf-deployed, compared to non-deployed military personnel, reporting depression (OR = 2.28,
95% CI 1.88-2.76). Overall heterogeneity, as indicated by I2, was high, at 75%. Stratification by
case-definition reduced the heterogeneity dramatically for the diagnostic interview subgroup (I2 =
0%) and the self-report physician diagnosis subgroup (I2 = 0%), though less dramatically for the
screening tool subgroup (I2 = 59%). The OR for the group of studies using a screening tool (OR =
2.71, 95% CI 2.23-3.31); (7, 8, 60-62, 65, 68) was higher than the OR for the groups of studies
using the diagnostic interview (OR = 1.75, 95% CI 1.47-2.01) (1, 6, 66, 67) or the self-report
physician diagnosis (OR = 1.82, 95% CI 1.49-2.24) (29, 63, 64).
The OR for the diagnostic interview subgroup indicated the odds of major depressive disorder in
Gulf-deployed, compared to non-deployed personnel, as all of the diagnostic interview studies
28
utilised DSM criteria for major depressive disorder rather than the more general overall outcome of
‘depression’.
Figure 2 Random effects meta-analysis of depression in Gulf War veterans compared to non-deployed military personnel
A meta-analysis stratified by adjustment of odds ratio (adjusted vs unadjusted) indicated little
differences on overall odds of depression between the groups (OR (adj subgroup) = 2.25, 95% CI
1.4-3.6 vs OR unadjusted subgroup = 2.57, 95% CI 2.2-3.0, forest plot not shown). Similarly, a
meta-analysis stratified by risk of bias (high vs low) indicated little differences in the overall odds of
depression between the groups (OR (high risk of bias) = 2.03, 95%CI 1.71-2.40 vs OR (low risk of
bias) = 2.30, 95% CI 1.75-3.04, forest plot not shown).
Dysthymia or chronic dysphoria meta-analysis
Five of the 14 studies summarised in Table 1 reported dysthymia (1, 6, 67) or chronic dysphoria (7,
8) as outcomes. The forest plot in Figure 3 indicates an overall odds ratio of similar magnitude to
depression; Gulf War veterans had over twice the odds of reporting dysthymia or chronic dysphoria
compared to non-deployed personnel (2.39, 95% CI 2.0-2.86). The overall heterogeneity between
studies was small (I2 = 0%). Consistent with MDD, studies utilising a diagnostic interview to
29
determine caseness (1, 6, 67) yielded an overall lower odds ratio (1.83, 95% CI 0.5-6.7) compared
to studies utilising screening tools (7, 8). The two studies utilising screening tools contributed
much greater weight to the calculation of the overall OR than the studies utilising diagnostic
interviews, due to the larger sample sizes in the studies using screening tools, although as was
also shown in Figure 2, screening tools generally produced higher odds ratios than more
methodologically rigorous structured diagnostic interviews.
Figure 3 Random effects meta-analysis of dysthymia or chronic dysphoria in Gulf War veterans compared to non-deployed military personnel
4.2.2 Key findings
Our systematic review and meta-analyses showed that Gulf War veterans were more than twice as
likely to experience depression compared with military personnel who were not deployed to the
Gulf War. The elevated odds of depression were statistically significant in 13 of the 14 studies that
were eligible to be included. This finding persisted when the meta-analysis was stratified by risk of
bias and by outcome measure. The overall odds of Gulf War veterans experiencing dysthymia or
chronic dysphoria compared to non-deployed personnel were also doubled. However, only five of
the 14 studies eligible to be included investigated dysthymia or chronic dysphoria, and three of the
five estimates were not statistically significant. In addition, two of the five studies were of chronic
dysphoria, rather than the DSM-diagnosed condition of dysthymia.
30
Table 1 Characteristics of eligible studies comparing prevalence of depression in Gulf War veterans and non-deployed military personnel
First author
Study design and study period
Sample Case definition and measure
Outcome GWV prevalence (%)
Comp group prevalence (%)
Comments and risk of bias assessment
Study samples using diagnostic interview to determine depression casenessWolfe 1999 (6)
Cross-sectional in-person structured diagnostic interview administered by trained clinicians; 1994 to 1996
Stratified random sample of two cohorts of US GW deployed veterans from New England Region (Fort Devens, FD; n = 148) and New Orleans (NO; n = 56) and a comparison group of air ambulance unit personnel deployed to Germany (G; n = 48) during the Gulf War period.
SCID non-patient edition using DSM-III-R criteria to assess current (1 month) MDD and Dysthymia prevalence
MDD 6.6 (FD) and 4.5 (NO)
0.0 Non-response bias assessed in study: yes, for demographics, health outcomesSignificant differences between respondents and non-respondents: yes, on demographics and health outcomes.Study participation rates: 62% (FD; 353 eligible); 38% (NO; 194 eligible); 85% (G; eligible numbers not reported).Psychological interview participation rates: 42% (FD), 30% (NO), 51% (G).Overall risk of bias: high
Dysthymia 3.6 (FD) and 4.8 (NO)
0.0
MDD unadjusted odds ratio combining FD and NO GWV, compared to German deployed comparison group = 6.3 (95% CI 0.4-108). Prevalence adjusted for stratification variables (health symptoms and gender).
Ikin 2004 (1)
Cross-sectional in-person structured diagnostic interview administered by trained psychologists; 2000 to 2003
All Australian GWV (n = 1381) and random sample of Navy, Army and Air force non-deployed active duty personnel (n = 1377) matched by age, gender and service type.
CIDI using DSM-IV criteria to assess post-Gulf MDD and Dysthymia prevalence
MDD 16.7 11.3 Non-response bias assessed in study: yes, for demographics, health outcomesSignificant differences between respondents and non-respondents: yes, on demographics; no on health outcomesParticipation rates: 81% GWV (1808 eligible); 57% non-GWV (2796 eligible).Psychological interview completion rates: 78% GWV; 51% non-deployed.Overall risk of bias: low
Dysthymia 0.4 0.3MDD OR = 1.6 (95% CI 1.3-2.0) adjusted for service type, rank, age, education, marital statusDysthymia OR = 1.4 (95% CI 0.3-7.2) adjusted for service type, rank, age.
Fiedler 2006
Cross-sectional telephone
Random sample (n = 967) of all US GWV and non-
CIDI-Short Form (SF) using DSM-IV
MDD 15.1 7.8 Non-response bias assessed in study: yes, for demographicsMDD OR = 2.07 (95% CI 1.50-2.85) adjusted for
31
First author
Study design and study period
Sample Case definition and measure
Outcome GWV prevalence (%)
Comp group prevalence (%)
Comments and risk of bias assessment
(66) structured diagnostic interview administered by trained interviewers; 2000 to 2001
deployed personnel (n = 784)
criteria to assess 12 month MDD prevalence
age, sex, rank, branch of service, duty status, education, marital status, ethnicity
Significant differences between respondents and non-respondents: yesParticipation rates: 59% GWV (1651 eligible); 51% non-deployed (1552 eligible).Psychological interview completion rate: 55% GWV; 43% non-deployed.Overall risk of bias: low
Toomey 2007 (67)
Cross-sectional in-person structured diagnostic interview, administered by trained interviewers; 1998 to 2001
Stratified random subsample of previous study (Kang 200031). US GWV (n = 1061) and non-deployed (n = 1128).
CIDI using DSM-IV criteria to assess Gulf –era onset (Jan 1991 to July 1993) MDD or Dysthymia; Current depression using BDI-II scores >13 indicating mild (14-19), moderate (17-29) and severe (30-63) depression
MDD 7.1 4.1 Non-response bias assessed in study: yes, for demographics, health outcomesSignificant differences between respondents and non-respondents: yes, on demographics; no on health outcomes.Psychological interview completion rates: 53% GWV (1996 eligible); 39% non-GWV (2883 eligible).Overall risk of bias: low
Dysthymia 0.04 0.0MDD OR = 1.81 (95% CI 1.03-3.19) adjusted for age, gender, ethnicity, education, duty type (active v reserve/guard), service branch and rank.Current depression as indicated by the BDI-II was significantly greater in GWV at the mild, moderate and severe levels compared to non-deployed personnel.
Samples using screening tools to determine depression casenessPerconte 1993 (60)
Cross-sectional in-person questionnaire administered by VAMC PTSD clinical team; study period not stated
Convenience sample of US reservists from Western Pennsylvania tri-state area GWV (n = 439), non-deployed comparison group (n = 126) and Europe-deployed group (n = 26).
BDI-I scores >10 indicating ‘minimal depression’ (common cutoff scores 32 for BDI-I are 10-18 indicating mild depression, 19-29 indicating moderate depression, 30-63 indicating severe depression)
Depression 26.9 16.7 Non-response bias assessed in study: noResponse rates: overall approximately 95% (620 eligible; denominators for GWV and non-deployed not provided)Overall risk of bias: high
Unadjusted OR = 1.84 (95% CI 1.1-3.1)
32
First author
Study design and study period
Sample Case definition and measure
Outcome GWV prevalence (%)
Comp group prevalence (%)
Comments and risk of bias assessment
Sutker 1993 (61)
Cross-sectional in-person questionnaire administered by VA staff; assessed 4 to 10 mths after return
Convenience sample of five US National Guard and Army Reserve Units in US state of Louisiana, with High war-zone stress (n = 110), low stress (105) and 60 non-deployed personnel
BDI-I scores >10 indicating ‘clinical depression’; war zone stress for GWV assessed using Operation Desert Storm (ODS) War-Zone Stress Exposure scale (ODS-SE)
Depression 36.0 (High Stress) and 13.0
(Low Stress)
12.0 Non-response bias assessed in study: noResponse rates: 70% GWV (306 eligible); non-GWV not providedOverall risk of bias: high
Unadjusted OR = 2.4 (95% CI 1.0-5.6)
Sutker 1995 (62)
Cross-sectional in-person questionnaire ; assessed within 1 year of their return from the Gulf region
912 US military personnel (GWV = 653, non-deployed = 259) drawn from overall sample of 1,423 Navy, Army, Air Force and Marine National Guard and Reserve Units mobilized for active duty. 511 of 1423 excluded from the analysis. Sample design not-reported.
BDI-I scores >10 indicating ‘clinical depression’
Depression 22.0 9.0 Non-response bias assessed in study: yes, for demographicsSignificant differences between respondents and excluded participants: noResponse rates: 64% overall (1423 eligible; denominators by serving gulf-war veteran status not provided)Overall risk of bias: high
Unadjusted OR = 2.9 (95% CI 1.8-4.5)
IOWA Persian Gulf Study Group 1997 (8)
Cross-sectional telephone interview; 1995 to 1996
Stratified random sample of US IOWA state Regular military (R) and National Guard/Reserve (NG/R) GWV (n = 1896) listing IOWA as home state and non-deployed personnel on active duty or activated during Gulf War (n = 1799) stratified by age, sex, ethnicity, rank and branch of service
PRIME-MD PHQ based on DSM-III-R criteria assessing 12 month symptoms of MDD and chronic dysphoria
MDD (R) 8.1 3.9 Non-response bias assessed in study: yes, for demographicsSignificant differences between respondents and non-respondents: yesResponse rates: 78% GWV (2421 eligible); 73% non-deployed (2465 eligible)Overall risk of bias: low
MDD (NG/R)
10.1 5.3
Dysphoria (R)
5.3 3.2
Dysphoria (NG/R)
8.4 4.0
MDD Combined Unadj OR = 2.1 (95% CI 1.6-2.7).Dysphoria Combined Unadj OR = 1.97 (95% CI 1.45-2.67)Prevalence rate differences adjusted for age, sex, race, branch of military and rank.
Goss Gilroy
Cross-sectional postal survey; 1997
All Canadian GWV (sea, land, air service; n = 3113)
PRIME-MD PHQ using DSM-III-R
MDD 18.9 (14.9b) 5.8 (4.9b) Non-response bias assessed in study: noChronic 10.7 (8.9b) 4.0 (4.3b)
33
First author
Study design and study period
Sample Case definition and measure
Outcome GWV prevalence (%)
Comp group prevalence (%)
Comments and risk of bias assessment
1998 (7) and sample of Canadian forces personnel eligible for active duty but non-deployed (n = 3439), matched on gender, age, regular/reserve status
criteria to assess current MDD and chronic dysphoria
Dysphoria (however, an assessment was made between GWV and non-deployed personnel, indicating no significant differences on demographics and confounding factors).Response rates: 73% GWV (4262 eligible); 60% non-deployed personnel (5699 eligible)Overall risk of bias: low
MDD Adjusted OR = 3.67 (95% CI 3.04-4.44) adjusted for rank and income.b = Gulf deployed with no other theatre experienceChronic Dysphoria Adjusted OR = 2.68 (95% CI 2.13-3.35) adjusted for rank , income, branch of service, age and education
Ishoy 2004 (65)
Cross-sectional in person questionnaire administered by physicians; 1997 to 1998
All Danish Gulf veterans (n = 686) and random sample of non-deployed comparison group matched on age, gender and profession (n = 231)
SCL-90-R current Depression dimension (factor scale, scores of 3+ on the depression dimension indicating depression)
Depression 11.0 3.9 Non-response bias assessed in study: no(however, an assessment was made between GWV and non-deployed personnel, indicating no significant differences on demographics but significant differences on health symptoms).Participation Rate: 84% (821 eligible) GWV; 58% non-deployed (400 potential participants)Overall risk of bias: high
Unadjusted OR = 3.0 (95% CI 1.5-6.2)
Kang 2009 (31)
Cross-sectional postal and telephone survey; 2004
Follow-up stratified random sample from previous study (Kang 200031) of US GWV (Navy, Army, Air Force, Marine; n = 6111) and non-deployed personnel frequency matched on gender, branch of service and service status (n = 3859)
PRIME-MD PHQ-9 using DSM-IV criteria to assess current MDD.
MDD 14.9 5.8 Non-response bias assessed in study: yes, on demographics, health outcome measures.Significant differences between responders and non-responders: yes, on demographics; no, on health outcomesResponse rates: 40% GWV (15,508 eligible); 27% non-deployed (14,494 eligible). Overall risk of bias: low
Unadjusted OR = 2.8 (95% CI 2.4-3.3). Adjusted RR = 2.34 (95% CI 2.03-2.70), adjusted for age, gender, race, BMI, cigarette smoking, rank, branch of service, unit component (active duty, national guard/reserve)
Samples using self-reported physician diagnosis to determine depression caseness
34
First author
Study design and study period
Sample Case definition and measure
Outcome GWV prevalence (%)
Comp group prevalence (%)
Comments and risk of bias assessment
Steele 2000 (29)
Cross-sectional telephone interview; 1998
Stratified random sample of US GWV residing in Kansas (n = 1545) and non-deployed comparison group (n = 435)
Self-reported physician diagnosis of depression in period from 1990 to 1998
Depression 12.0 7.0 Non-response bias assessed in study: yes, demographicsSignificant differences between responders and non-responders: yesParticipation rate: 63% overall (3,138 eligible).Response rates: 93% GWV; 88% non-deployedOverall risk of bias: high
Adjusted OR = 1.85 (95% CI 1.22-2.81) adjusted for age, sex, income and education
Gray 2002 (63)
Cross-sectional postal survey; 1997 to 1999
Sample of all US Gulf War-era Seabees (members of US Naval Mobile Construction Battalions; n = 3831) Gulf-era Seabees deployed elsewhere (n = 4933) and Gulf-era non-deployed Seabees(n = 3104)
Self-reported physician diagnosis of depression diagnosed since 1991; and in past 12 months
Depression 7.7 4.6 Non-response bias assessed in study: yes, demographics, health outcomesSignificant differences between responders and non-responders: yes, on select demographics; yes on health outcomes.Response rates:63% overall (18,945 eligible); 70% of those located participated.Overall risk of bias: high.
Adjusted OR = 1.77 (95% CI 1.41-2.27)adjusted for age, gender, active-duty/reserve status, race/ethnicity, current smoking and current alcohol use
McCauley 2002 (64)
Cross-sectional telephone interview; 1998 to 1999
Random sample of three groups of US Army or National Guard veterans living in five US states (OR, WA, CA, NC and GA) and 1. serving within 50km of Khamisiyah Iraqi munitions site (KHAM GWV n = 653), 2. non-Khamisiyah deployed (GWV other deployed n = 610) and 3. non-deployed personnel (n = 516)
Self-reported physician diagnosed MDD with hospitalization
MDD 1.7 0.6 Non-response bias assessed in study: yes, demographicsSignificant differences between responders and non-responders: yesResponse rates: 71% KHAM GWV; 38% non-deployedOverall risk of bias: high
Adjusted OR (deployed v non-deployed) = 5.1 (95% CI 1.5-32.1) adjusted for age, gender, race and region of residence
35
4.3 Depression in Afghanistan/Iraq War veterans compared to non-deployed military personnel
4.3.1 Results
Thirteen studies met the eligibility criteria (69-81). Of these, two papers (Cabrera 2007 (72),
Nguyen 2013 (76)) were excluded from the meta-analysis because their samples completely
overlapped with a larger sample of another included study (Hoge 2004 (74) and Wells 2010 (81),
respectively). Also, one study (Vasterling 2006) (80), only reported mean (standard error) scores
on the Center for Epidemiologic Studies Depression Scale (CES-D) scale rather than binary data
needed to calculate ORs, and so was excluded from the meta-analysis. Of the remaining ten
studies, two (Shen 2012, (78) and Wells 2010 (81)) reported adjusted ORs for subgroups (e.g.
Wells 2010 reported separate ORs for males and females). We combined subgroup ORs within
these studies so that each study contributed only one effect estimate to the meta-analysis.
Main analysis depression in Afghanistan/Iraq War veterans
Based on a random-effects meta-analysis of ten studies, there was an increased odds of
depression in deployed Afghanistan/Iraq War veterans compared with non-deployed personnel
(OR 1.58, 95% CI 1.14 to 2.17) (Figure 4). However, there was a very high amount of statistical
heterogeneity (I2 = 98%), so the meta-analytic effect should be interpreted with caution in terms of
the degree of elevated odds. Across studies, the magnitude of the ORs ranged from 0.94 to 3.08.
36
Figure 4 Random effects meta-analysis of depression in Afghanistan/Iraq War veterans compared to non-deployed military
Subgroup and sensitivity analyses depression in Afghanistan/Iraq War veterans
Sources of heterogeneity were assessed by subgroup analyses according to the outcome measure
used (ICD-9 diagnosis, screening instrument PHQ-9, or other (e.g. abbreviated screening
instrument e.g. PHQ-2)), service type of veterans (all services, National Guard only, or Air Force
only), risk of bias, and adjustment of ORs.
Stratification by type of outcome measure used to diagnose depression identified important
differences in subgroups (Figure 5). The meta-regression identified a statistically significant
difference between these three subgroups, with a relationship suggesting that the more rigorous
the outcome measure, the higher the OR (P = 0.05).
Figure 5 Random-effects meta-analysis of depression in Afghanistan/Iraq War veterans compared to non-deployed military personnel, subgrouped by type of outcome measure
37
Odds Ratio
In contrast, stratification by type of service did not reveal important differences between subgroups
(Figure 6). Results of the meta-regression suggested no statistically significant association
between type of service and magnitude of OR (P = 0.45).
Figure 6 Random-effects meta-analysis of depression in Afghanistan/Iraq War veterans compared to non-deployed military personnel, subgrouped by type of service
Stratification by risk of bias did not reveal important differences between subgroups (Figure 7).
Results of the meta-regression suggested no statistically significant association between risk of
bias and magnitude of OR (P = 0.99).
38
Figure 7 Random-effects meta-analysis of depression in Afghanistan/Iraq War veterans compared to non-deployed military personnel, subgrouped by risk of bias
Stratification by adjustment of ORs did not reveal important differences between subgroups (Figure
8). Results of the meta-regression suggested no statistically significant association between
adjustment of OR and magnitude of OR (P = 0.08).
39
Figure 8 Random-effects meta-analysis of depression in Afghanistan/Iraq War veterans compared to non-deployed military personnel, subgrouped by adjustment of odds ratios
Sensitivity analyses indicated that the overall OR did not change after excluding any individual
study, and the statistical significance did not change (Table 2). However, in all sensitivity analyses,
statistical heterogeneity remained high (i.e. I2 > 94%).
Table 2 Sensitivity analyses excluding each study one by one for studies of depression in Afghanistan/Iraq War veterans compared to non-deployed military personnel
Excluded study Pooled OR LCI 95% HCI 95% Cochran Q Chi2 P-value I2 (%)
Bray 2006 1.67 1.21 2.31 335.12 0.00 98
Bray 2009 1.65 1.18 2.32 388.84 0.00 98
Bray 2011 1.66 1.19 2.31 341.92 0.00 98
Hoge 2004 1.60 1.13 2.26 427.70 0.00 98
Hoge 2006 1.53 1.04 2.25 428.67 0.00 98
Kline 2010 1.48 1.06 2.07 429.31 0.00 98
Peterson 2010 1.54 1.09 2.17 433.11 0.00 98
Shen 2012 1.44 1.14 1.81 129.60 0.00 94
Vanderploeg 2012 1.61 1.16 2.24 432.03 0.00 98
Wells 2010 1.58 1.10 2.27 426.23 0.00 98
40
The funnel plot was not asymmetrical (Figure 9), and the Egger test was not statistically significant
(P = 0.30). (Note: FE MA, Fixed effects meta-analysis)
Figure 9 Funnel plot for the random-effects meta-analysis of depression in Afghanistan/Iraq War veterans compared to non-deployed military personnel
Comparison of summary odds ratio of depression in Gulf War veterans vs Afghanistan/Iraq War veterans
The P-value for the test for equality of the summary OR of depression in Gulf War veterans’ meta-
analysis (OR = 2.28, 95% CI 1.88-2.76; I2 75%) and summary OR of depression in
Afghanistan/Iraq War veterans’ meta-analysis (OR 1.58, 95% CI 1.14 to 2.17; I2 = 98%) was 0.055.
This is suggestive of depression in Gulf War veterans being higher than in Afghanistan/Iraq War
veterans compared with their non-deployed military comparison groups respectively, but the
difference between the two summary ORs narrowly missed statistical significance.
41
Table 3 Characteristics of eligible studies comparing prevalence of depression in Afghanistan/Iraq War veterans and non-deployed military personnel
Citation Study design and period
Sample Depression case measure and definition
Outcome Afghanistan/Iraq deployed prevalence
(%)
Comp group prevalence (%)
Comments and overall risk of bias assessment
Wells 2010 (81)
Cohort study randomly selected from all US military serving in October 2000 (Millennium Cohort Study, 1st/baseline panel 2001-3, follow up 2004-6). Internet or paper survey in 2004-06
40,219 cohort participants who completed both questionnaires and met inclusion criteria (n=55021). Exclusion criteria included indicators of depression at baseline, deployed prior to/while completing 2001-3 questionnaire, incomplete data. Participants in 1st panel who deployed to Iraq or Afghanistan Wars between questionnaire surveys with (n=225) and without (n=92) combat exposure and those who did not deploy (n=872)
PHQ defined new onset depression: (1) endorsed having depressed mood/anhedonia, and (2) responded “> half the days” or “nearly every day” to ≥5 of 9 PHQ items in past 2 weeks, suicidal ideation if present at all. Combat exposed if ≥1 specified combat experience in past 3 years
New onset depressionMales
Combat exposed 5.7Non-combat exposed 2.3
3.9 1st panel response rate was 31%, but published investigations suggest sample is representative of military personnel including deployers, reporting is reliable, response bias is minimal.Overall risk of bias: Low
Females Combat exposed 15.7Non-combat exposed 5.1
7.7
Deployed with combat exposure vs not deployed: males OR 1.32 (95% CI 1.13-1.54); females OR 2.13 (1.70-2.65) adjusted for deployment status, birth year, education, marital, smoking, alcohol, baseline PTSD symptoms/diagnosis, rank, service component and branch, occupation.Deployed without combat exposure vs not deployed: males adj OR 0.66 (0.53-0.83) females OR 0.65 (0.47-0.89).Females (n=10,178) were 25.3% of cohort sample
Kline 2010 (75)
Cross sectional anonymous self-administered survey of National Guard (NG) undergoing Iraq pre-deployment medical assessments; 2007-08
New Jersey (NJ) NG (n=2543) undergoing pre-deployment for Iraq: two groups; ≥1 prior OEF/OIF deployment since 2001 (n=625) and no prior OEF/OIF deployments (n=1910). Approx. 14.5% had served in conflicts other than OEF or OIF. De-identified Iraq pre-deployment health data on all 2995 NG from NJ Department of Military and Veterans Affairs (DMAVA)
Major depression and any depression (major depression or depression not otherwise specified) defined using PHQ-9.Prescribed use of antidepressant past 12 months
Major depression
5.1 2.0 Response rate =2665/2995 (89.0%). Sample approx. 50% of NJ NG force. NJ NG included more Hispanics, fewer non-Hispanic whites than a national NG sample, though few differences in current data on race. Depression 0.8% in DMAVA NG health assessment data vs 3.4% in study survey. Non response bias not assessed due to survey anonymityOverall risk of bias: low
Previously deployed to war in Iraq or Afghanistan vs non-OEF/OIF deployed group OR=3.07 (95% CI 1.81-5.19) adjusted for age, sex, race/ethnicity, education, income, marital status, and military deployment other than OEF/OIF.Any depression 6.6 2.3Adj OR=2.94 (95% CI 2.09-4.13)Prescribed use antidepressant
9.8 2.8
Adj OR=3.54 (95% CI 2.35-5.33)
42
Citation Study design and period
Sample Depression case measure and definition
Outcome Afghanistan/Iraq deployed prevalence
(%)
Comp group prevalence (%)
Comments and overall risk of bias assessment
Vander-ploeg 2012 (79)
Cross sectional anonymous online survey of currently active Florida NG; 2009-10
10,400 currently active Florida NG invited, 4005 responded. 3098 in final sample (1443, 46.6% deployed to Afghanistan/Iraq, 1655 not deployed)
Major depression during past month using PHQ-9 if DSM-IV criteria met and work, home or interpersonal impairment reported at ‘very difficult’ level
Major depression
3.3 1.0 Response rate 41.3%Non response bias not assessed due to survey anonymity. No information on Florida NG cohort available.Overall risk of bias: high
OR 0.94 (95% CI 0.29-3.07) adjusted for demographics, pre deployment psychological trauma or traumatic brain injury (TBI), and deployment related factors including combat, physical injuries, potentially traumatic combat experiences, blast exposure and TBI. In deployed with a single probable mild TBI OR 2.55 (95% CI 1.40-4.64) and with multiple TBIs OR 4.73 (1.61-13.89)
Shen 2012 (78)
Random sample of approximately 25% of all active duty US personnel from all services between 2001 and 2006, by combining databases (US defence and inpatient and outpatient health information from all civilian and military health providers)
678,382 US personnel from all services: 333,548 (49%) Army, 98,524 (14%) Marines, 134,015 (20%) Navy 112,295 (17%) Air Force. Deployment groupings were Afghanistan/Iraq deployed, deployed at other known locations under OIF/OEF (such as Kuwait, Qatar, Saudi Arabia, Turkey), deployed to classified or unknown locations
Enlisted persons diagnosed with major depression ICD-9 code either 296.2 or 296.3 anytime between 2001 and 2006
ICD-9 coded major depression
Did not give overall percentages for all group, gave by service type and by deployment location
Sample was representative of the US Armed Forces active duty enlisted populationOverall risk of bias: low
Peterson 2010 (77)
Cross-sectional survey post deployment of US personnel deployed to Iraq (combat zone) or Qatar (non combat zone) based on de-identified data
Participants were active duty US Air Force non combatant personnel deployed to Iraq (n=4,408) or Qatar (“non deployed” comparison group) (n=959) who completed a PDHA; 2005-2007
PHQ-2 to measure symptoms of depressed mood or anhedonia; ≥1 positive response considered screen positive for depression
Depression screen positive
9.9 5.4 Previous reports showed high completion rates for Health Assessments but response rate not reported/applicable. Demographic characteristics of groups similar in gender, age, marital, and military
OR 1.90 (95% CI 1.40-2.59). Not stated whether OR was adjusted for possible confounding factors
43
Citation Study design and period
Sample Depression case measure and definition
Outcome Afghanistan/Iraq deployed prevalence
(%)
Comp group prevalence (%)
Comments and overall risk of bias assessment
US Department of Defense (DoD) Post-Deployment Health Assessment (PDHA) surveys; 2005-2007
grade, but average time in service unavailable. Deployment length = 4m. Generalisability for other services could be limited.Overall risk of bias: high
Hoge 2004 (74)
Anonymous cross-sectional survey phase of a longitudinal study; 2003
Samples from an Army combat infantry brigade pre deployment to Iraq (n=2530); an infantry brigade of same division 6m post deployment to Afghanistan (n = 1962); an Army infantry brigade 8m post Iraq deployment (n = 894); and Marine Corp units 6m post Iraq deployment (n = 815)
Questionnaires administered 3-4m after return to US. PHQ-9 to assess major depressive disorder in past month using 2 outcome variables: broad screening definition using DSM-IV criteria; and conservative (strict) screening definition also requiring substantial functional impairment or a large number of symptoms
Depression (broad)
15.2 (Army post Iraq)14.2 (Army post Afghanistan)14.7 (Marines post Iraq)
11.4 (Army pre deployment Iraq)
Response rate: (defined as completion of any part of survey) of 98% among the 58% of combined samples available to attend study briefings
Demographic characteristics of sample similar to the general, deployed, active duty infantry population
Overall risk of bias: low
Adj OR Army post Iraq 1.40 (95% CI 1.12-1.76),Adj OR Army post Afghanistan 1.29 (1.07-1.54), adj for age, rank, education, marital, race/ethnicityUnadj OR Marines post Iraq 1.34 (1.06-1.70),vs Army pre deployment to IraqDepression (strict) 7.9 (Army post Iraq)
6.9 (Army post Afghanistan)7.1 (Marines post Iraq)
5.3 (Army pre deployment Iraq)
Adj OR Army post Iraq 1.53 (95% CI 1.12-2.08),Adj OR Army post Afghanistan 1.33 (1.03-1.71), adj for age, rank, education, marital, race/ethnicity,Unadj OR Marines post Iraq 1.37 (0.99-1.90)All vs Army pre deployment to Iraq
Hoge 2006 (73)
Analysis of Defence Medical Surveillance System (DMSS) database for data on all US Army soldiers and Marines who completed a routine PDHA between 1 May
Army soldiers and Marines post deployment to Iraq (n=222,620), Afghanistan (n=16,318), and other locations, e.g. Bosnia, Kosovo (n= 64,967) who completed the routine PDHA
Modified PHQ-2 measures depressed mood and anhedonia; ≥1 positive response considered screen positive for depression. Administered using electronic device or paper survey, pre deployment or within 1-2 weeks of return home. Referral for mental
PHQ-21 item
PHQ-22 items
Iraq 4.5*
Afghanistan 2.5
Iraq 1.6*
Afghanistan 1.0
1.9
0.8
Demographics of study population and study findings using survey and electronic versions of the PDHA were similar so combined. 18% of study population did not have a PDHA; similar but more likely to be active duty Marines. Some demographic differences
* Iraq deployment associated with depression compared with Afghanistan and other locations after controlling for demographics. Afghanistan deployment associated with depression compared with
44
Citation Study design and period
Sample Depression case measure and definition
Outcome Afghanistan/Iraq deployed prevalence
(%)
Comp group prevalence (%)
Comments and overall risk of bias assessment
2013 to 30 April 2004
health problem, mental health care utilisation, attrition assessed during 12m follow up period.
other location after controlling for demographic factors. OR not given in paper.
between deployment groupsOverall risk of bias: low
Shen 2012 (78)
Database analysis on several combined sources from US Defense Manpower Data Center and TRICARE, to obtain active duty personnel, demographic service, mental and related health information
678 382 unique active duty personnel serving between 2001 and 2006, approx 49% Army, 14% Marine, 20% Navy, 17% Air Force. Four groups: Not deployed under OEF/OIF; Deployed- Iraq or Afghanistan; Other known locations under OEF/OIF, e.g. Kuwait, Qatar, Turkey; Classified/ unknown locations
ICD-9 diagnosis of major depression between 2001 and 2006. Also considered 3 deployment lengths (data not shown)
ArmyMarinesNavyAir Force
5.13.85.83.5
1.71.22.23.1
Sample 25% and representative of US Armed Forces active duty population, Captured diagnoses from in/outpatient settings, civilian and military providers.Undertook sensitivity analyses. OR adj for demographic (gender, race/ethnicity, marital age) and service characteristicsOverall risk of bias: low
Army Adj OR 3.52 (95% CI 3.21-3.86)Marines Adj OR 4.51 (3.66-5.57)Navy Adj OR 3.25 (2.50, 4.22)Air Force Adj OR 1.45 (1.22, 1.72)Reference group is not deployed under OIF/OEFOEF-OIF deployed females - Army 11% Marines 3%, Navy 13%, Air Force 15% and similar to Not deployed.
Bray 2006 (69)
Cross sectional self-report anonymous DoD survey of Health Related Behaviour Among Active Duty Military Personnel (HRBS); 2006 (9th data point in series since 1980)
Sample of all eligible active duty US military personnel (excluded recruits, academy students, AWOL, or had permanent change of station). n=16,146 (3639 Army, 4627 Navy, 3356 Marine, 4524 Air Force)
40,436 US active duty service members were sampled from installations or at remote locations with 28,546 completing the surveys (5927 Army, 6637 Navy, 5117 Marine Corps, 7009 Air Force and 3856 Coast Guard), Data were weighted to represent all active duty personnel
Index of Need for Further Depression Evaluation 3-item Version A Burnam depression developed based on 2 items from CES-D and 1 item from the Diagnostic Interview Schedule based on reports of extended period of depression symptoms primarily in past 12 months
Need for Further Depression Evaluation
22.32 22.48 Overall response rate: 51.8%.Participants were selected to represent men and women in all pay grades of the active force worldwide. Data were weighted to represent all active duty personnel in the analyses.Overall risk of bias: low
Unadjusted OR 95% CI in OEF/OIF veterans vs did not serve any operation
45
Citation Study design and period
Sample Depression case measure and definition
Outcome Afghanistan/Iraq deployed prevalence
(%)
Comp group prevalence (%)
Comments and overall risk of bias assessment
Bray 2009 (70)
Cross sectional self-report anonymous Health Related Behaviour Among Active Duty Military Personnel questionnaire survey; 2008 (10th data point in series since 1980, included active duty Coast Guard for first time)
Sample of all eligible active duty US military personnel (excluded recruits, academy students, AWOL or incarcerated personnel). n=28,546; (5927 Army, 6637 Navy, 5117 Marine, 7009 Air Force, 3856 Coast Guard)
Index of Need for Further Depression Evaluation 3-item Version A Burnam depression developed based on 1 item from the CES-D and 2 items from the Diagnostic Interview Schedule based on reports of both current and extended periods of depression in past 12 months
Need for Further Depression
EvaluationAllArmyNavyMarinesAir ForceDoD ServicesCoast Guard
21.6426.2919.4225.1313.5221.63
26.86
20.4222.2921.9125.7814.0520.54
17.71
Overall response rate: 71.6%.Participants were selected to represent men and women in all pay grades of the active force worldwide. Data were weighted to represent all active duty personnel in the analyses.Overall risk of bias: low
Unadjusted OR 95% CI combat deployed in past year and ever served in OEF/OIF vs not combat deployed past year
Bray 2011(71)
Anonymous cross-sectional DoD web-based survey of Health Related Behaviour Among Active Duty Military Personnel (HRBS), 2011
Stratified random sample of all members of the Army, Navy, Marine Corps, Air Force, and Coast Guard who were non-deployed and on active duty at time of the study. Eligible sample size was 154,011 from the DoD services and 14,653 from the USCG. Stratified random sample DoD by service, gender, and pay grade, US CG by work setting (ashore, afloat, air), gender, and pay grade. Eligible respondent 34416 DoD ( Army 6,932, Navy 7,571, Marine Corps 8,339, Air Force 11,574) USCG 5461
Two questions “I felt depressed” and “I felt sad” past week scored on 5-point scale, ranging from “Never” to “5-7 days. Scores coded and categories as high/low depression level past week
High level depression past week
ArmyNavyMarinesAir ForceCGAll services
Combat deployed since Sept 11, 2001 served in OIF/OEF
47.8 (1.1)46.0 (SE 1.2)47.7 (SE 1.2)33.4 (SE 0.7)42.8 (SE 2.7)44.1 (SE 0.6)
Not combat deployed since Sept 11
39.7 (SE 1.5)40.9 (SE 1.5)50.1 (SE 1.3)29.2 (SE 0.8)31.8 (SE 1.1)38.1 (SE 0.7)
Response rate: DoD 22%, CoastGuard 37%Overall risk of bias: low
46
4.3.2 Key findings
Based on a random-effects meta-analysis, there was an increased odds of depression in
deployed Afghanistan/Iraq War veterans compared with non-deployed personnel. However,
there was a very high amount of statistical heterogeneity, so the meta-analytic effect should
be interpreted with caution in terms of the actual level of the increased odds.
Stratification by type of outcome measure used to diagnose depression identified a
statistically significant difference between these three subgroups (ICD-9, screening
instrument or other), with a relationship suggesting that the more rigorous the outcome
measure, the higher the OR. Other subgroup analyses did not identify a statistically
significant difference between subgroups and the high heterogeneity persisted.
The difference in the summary OR for depression in Gulf War veterans compared with
Afghanistan/Iraq War veterans from the meta-analyses narrowly missed statistical
significance.
47
4.4 PTSD in Gulf War veterans compared to non-deployed military personnel
4.4.1 Results
Twenty studies met the eligibility criteria (1, 6-9, 17, 29, 31, 60, 63, 64, 66, 67, 82-88). Two
studies (9, 88) reported prevalence data on the same cohort, but Unwin 1999 reported an
OR for males while Unwin 2002 reported an OR for females. We combined these gender-
specific ORs so that only one effect estimate for the Unwin studies contributed to the meta-
analysis. Three studies (8, 17, 87) reported separate ORs for regular and reservist veterans.
We included each subgroup as a separate comparison in the meta-analysis.
Main analysis
Based on a random-effects meta-analysis, there was an increased odds of PTSD in
deployed compared with non-deployed veterans (OR 3.39, 95% CI 2.79 to 4.13) (Figure 10).
There was a moderate amount of statistical heterogeneity (I2 = 53%).
48
Figure 10 Random-effects meta-analysis of PTSD in Gulf War veterans and non-deployed military personnel
Subgroup and sensitivity analyses
Sources of heterogeneity were assessed by subgroup analyses according to the outcome
measure used (structured interview versus screening instrument versus self-reported
physician diagnosis), type of service (regular versus reservist), risk of bias and adjustment of
ORs.
Stratification by type of outcome measure did not reveal important differences between
subgroups (Figure 11). Results of the meta-regression suggested no statistically significant
association between type of outcome measure and magnitude of OR (P = 0.71).
49
Odds Ratio
Figure 11 Random-effects meta-analysis of PTSD in Gulf War veterans and non-deployed military personnel, subgrouped by type of outcome measure
Stratification by regular versus reservist service did not reveal important differences between
subgroups (Figure 12). Results of the meta-regression suggested no statistically significant
association between type of service and magnitude of OR (P = 0.45).
Figure 12 Random-effects meta-analysis of PTSD in Gulf War veterans and non-deployed military personnel, subgrouped by type of service
50
Stratification by risk of bias revealed important differences between subgroups (Figure 13).
Results of the meta-regression suggested a larger OR in the high risk of bias studies (P =
0.003).
Figure 13 Random-effects meta-analysis of PTSD in Gulf War veterans and non-deployed military personnel, subgrouped by risk of bias
51
Stratification by adjustment of ORs did not reveal important differences between subgroups
(Figure 14). Results of the meta-regression suggested no statistically significant association
between adjustment of OR and magnitude of OR (P = 0.25).
Figure 14 Random-effects meta-analysis of PTSD in Gulf War veterans and non-deployed military personnel, subgrouped by adjustment of odds ratios
Sensitivity analyses indicated that the overall OR did not vary after excluding any individual
study, and the statistical significance did not change (Table 4).
52
Table 4 Sensitivity analyses excluding each study one by one for studies of PTSD in Gulf War veterans compared to non-deployed military personnel
Excluded study Pooled OR LCI 95% HCI 95% Cochran Q Chi2 P-value I2 (%)
Black 2004 (regular) 3.43 2.80 4.20 42.10 0.00 55
Black 2004 (reservist) 3.51 2.87 4.29 39.61 0.00 52
Fielder 2006 3.38 2.75 4.14 42.13 0.00 55
Goss Gilroy 1998 3.40 2.75 4.20 42.28 0.00 55
Gray 2002 3.34 2.71 4.11 41.03 0.00 54
Holmes 1998 3.38 2.77 4.13 42.18 0.00 55
Ikin 2004 3.36 2.73 4.14 41.74 0.00 54
Iowa 1997 (regular) 3.42 2.79 4.19 42.20 0.00 55
Iowa 1997 (reservist) 3.47 2.84 4.24 40.84 0.00 53
Kang 2003 3.43 2.69 4.38 41.60 0.00 54
McCauley 2002 3.28 2.72 3.95 35.90 0.01 47
Murphy 2006 3.51 2.94 4.18 31.89 0.03 40
Perconte 1993 3.37 2.75 4.12 41.94 0.00 55
Pontius 1992 3.35 2.75 4.09 41.03 0.00 54
Smith 2009 3.52 2.90 4.27 37.77 0.01 50
Steele 2000 3.35 2.73 4.10 41.46 0.00 54
Stretch 1996 (regular) 3.24 2.67 3.93 35.84 0.01 47
Stretch 1996 (reservist) 3.32 2.70 4.07 40.34 0.00 53
Toomey 2007 3.31 2.71 4.05 40.14 0.00 53
Unwin 1999 3.47 2.78 4.33 41.13 0.00 54
Wolfe 1999 3.38 2.77 4.13 42.10 0.00 55
The funnel plot was not asymmetrical (Figure 15), and the Egger test was not statistically
significant (P = 0.49), indicating that publication bias was not present. (Note: FE MA, Fixed effects
meta-analysis)
53
Figure 15 Funnel plot for the random-effects meta-analysis of PTSD in Gulf War veterans and non-deployed military personnel
54
Table 5 Characteristics of eligible studies comparing prevalence of PTSD in Gulf War veterans and non-deployed military personnel
First author
Study design and period
Sample Case definition and measure
Outcome GWV prevalence (%)
Comp group prevalence (%)
Comments and risk of bias assessment
Pontius1992 (85)
Mental health screening and outreach programs conducted by a clinical team during which a survey was completed by participants
Sample of deployed and non-deployed Army, Navy and Marine reserve units in the Western Pennsylvania, Eastern Ohio and contiguous West Virginia area following Gulf War (n=537)
Mississippi Scale for Combat Related PTSD (M-PTSD)-Revised. Cut off score ≥89
PTSD 16.15 (n=57) 2.13 (n=2) Non-response bias assessed in study: not assessed.Self-selection (non-random) of participantsOverall risk of bias: high
Study population composition: 87.5% males
Males and females included in the analysis.
Perconte 1993 (60)
Cross-sectional in-person questionnaire administered by Veterans Affairs Medical Centre (VAMC) PTSD clinical team; study period not stated
Convenience sample of US reservists (Army, Navy and Marine) from Western Pennsylvania tri-state area; GWV (n = 439), non-deployed (ND) (n = 126) and Europe-deployed (n = 26) group
M-PTSD. Cut off score ≥89
PTSD 15.53 3.97 (ND) Response rates: overall approximately 95% (620 eligible; denominators for GWV and non-deployed not provided)Non-response bias assessed in study: noSmall number of cases of PTSD in comparison groups, n=5 ND, n=1 Europe deployed, n=68 in GWVOverall risk of bias: high
88.2% of participants were males. Males and females included in the analysis.
Stretch 1996 (87)
Anonymous cross sectional postal questionnaire; no date stated
Active duty and reserve veterans from Pennsylvania and Hawaii who deployed either to the Persian Gulf as a result of ODS in the Army, Navy, Air Force and Marines or did not deploy anywhere (non-deployers)
PTSD algorithm utilising selected items from Impact of Event Scale and Brief Symptom Inventory to denote possible risk of PTSD based on DSM-III-R diagnostic criteria
PTSD (Active) 8.0 (57/715) 1.3 (21/1576) Response rate: 31% (of potential population, n=1524 deployed to Persian Gulf, non-deployed anywhere during Operation Desert Storm (n=2512)Non-response bias: Not assessedOverall risk of bias: high
PTSD (Regular)
9.2 (70/761) 2.1 (20/945)
Proportion of males and females in the sample not stated
55
First author
Study design and period
Sample Case definition and measure
Outcome GWV prevalence (%)
Comp group prevalence (%)
Comments and risk of bias assessment
Wolfe 1999 (6)
Cross-sectional in-person structured diagnostic interview administered by trained clinicians; 1994-1996
Stratified random sample of two cohorts of US GW deployed veterans from New England Region (Army) (Fort Devens, FD, n = 148) and New Orleans, NO (all military branches (n = 56) and a comparison group of air ambulance unit personnel deployed to Germany (n = 48) during the Gulf War
CAPS (structured clinical interview) to assess clinical levels of PTSD
35-item M-PTSD modified for Desert Storm personnel
PTSD (current)
5.4 (FD) and 7.2 (NO)
0.0 Participation rates: 62% (FD; 353 eligible); 38% (NO; 194 eligible); 85% (Germany; not reported). Psychological interview participation rates: 42% (FD), 30% (NO), 51% (Germany)Non-response bias assessed in study: yes in both demographics and health outcomesSignificant differences between respondents and non-respondents: yes in both demographics and health outcomesOverall risk of bias: high
PTSD (lifetime)
6.5 (FD) and 8.2 (NO)
0.0
Prevalences were adjusted for stratification variables (gender and reported health symptoms).
FD - GWV 91.7% males, NO 78.8% males, and Germany 87.5% males.Males and females included in the analysis.
IOWA Persian Gulf Study Group 1997(8)
Cross-sectional telephone interview; 1995 to 1996
Stratified random sample of US Iowa state Regular military (R) and National Guard/Reserve (NG/Res) GWV (n = 1896) and active duty or activated non-deployed personnel (n = 1799) stratified by service, age, sex, ethnicity and rank
PTSD Checklist –Military (PCL-M). Cut-off score ≥50
PTSD (R) 1.9 0.7 Response rates: 78% GWV (2421 eligible); 73% non-deployed (2465 eligible)Non-response bias assessed in study: yes, for demographics, significant differences between responders and non-respondersOverall risk of bias: low
PTSD (NG/R) 2.0 1.1
Prevalence rate difference adjusted for age, sex, race, branch of military and rank:Total GW vs non-deployed 0.9 (95% CI -0.3, 1.5)R vs non deployed 0.9 (05% CI -0.1, 1.9)NG/R vs non-deployed 0.9 (95% CI 0.0, 1.7)
8.9% of eligible sample were female. Males and females included in the analysis
Holmes Cross sectional Air National Guard from A version of the M-PTSD 6.8 1.7 Response rate: 46% of original
56
First author
Study design and period
Sample Case definition and measure
Outcome GWV prevalence (%)
Comp group prevalence (%)
Comments and risk of bias assessment
1998(82)
postal survey; 1992
one Unit members activated into the Air Force in preparation for GW (n=517) and remaining unit members not deployed and remained in US (n=497)
M-PTSD
Categorical cut off score ≥89
sample (58.5% deployed, 41.5% non-deployed)Non-response bias: assessed, higher response from deployed and from higher ranks.Overall risk of bias: high
87% participants male. Males and females included in the analysis.
Goss Gilroy 1998 (7)
Cross-sectional postal survey; 1997
All Canadian GWV (sea, land, air service; n = 3113) and comparison group sample of Canadian forces personnel eligible for active duty but ND (n = 3439), matched on gender, age, military duty (regular/reserve status)
PCL-M to assess PTSD symptoms and health care provider-diagnosed PTSD
PTSD (symptoms)
2.5a (2.1b) 1.2a (0.6b) Response rates: 73% GWV (4262 eligible); 60% non-deployed personnel (5699 eligible)Non-response bias assessed in study: noA comparison between GWV and non-deployed personnel indicated no significant differences on socio-demographic and possible confounding factors.Overall risk of bias: low
PTSD (provider diagnosed)
3.1a (2.0b) 1.4a (0.5b)
GWV vs comparison group:PTSD (symptoms) adj OR = 2.69 (95% CI 1.69-4.26) adj for incomePTSD (provider diagnosed) adj OR = 3.34 (95% CI 2.13-5.26) adj for income, service
a with other theatre experience, b with no other theatre experienceOf participants 93.9% GWV were male, 94.2% comparison group were male. Males and females included in the analysis.
Unwin 1999(9)
Cross sectional postal survey; 1997-1998
Random stratified sample of UK service personnel (Army, Navy and Royal Air Force) deployed to GW (n=2735), Bosnia (n=2393) and non-deployed Era cohort (n=2422). Stratified by service, sex, age, service status, rank and fitness (army, air force)
Symptoms of Post-Traumatic Stress Reaction (PTSR) taken from the M-PTSD
PTSR 13.2 4.7 (Bosnia)4.1 (Era)
Response rate: 65.1% (valid responses); 70.4% GW, 61.9% Bosnia, 62.9%, Era cohort.Non-response bias assessed (tracing a randomly selected sample of 100 GW, 50 Bosnia, 50 Era participants, who were non-responders after two mailings)-significant differences between responders and non-responders: (Yes, demographics (by age and
PTSR OR (GW vs Era) = 2.7 (95% CI 2.1-3.6)PTSR OR (GW vs Bosnia) = 2.3 (95% CI 1.7-3.2)
GW cohort 92.4% males, Bosnia 91.3% males, Era 92.7% males. Males only included in the analysis.
57
First author
Study design and period
Sample Case definition and measure
Outcome GWV prevalence (%)
Comp group prevalence (%)
Comments and risk of bias assessmentserving status); health outcomes: noOverall risk of bias: low
Steele 2000(29)
Cross-sectional telephone interview; 1998
Stratified random sample of US Kansas resident GWV (n = 1545) and non-deployed (ND) comparison group (n = 435). Consisting of Army, Air Force, Navy, Marines and Coast Guard branches
Self-reported physician diagnosed PTSD new onset 1990-98 in period from 1990 to 1998
PTSD 6.0 1.0 Participation rate: 65% overall, and 93% GWV vs 88% non-deployed eligible participated. 15% GWV’s military records indicated they had not served in the GW but reported that they had (excluded from analyses). Small number PTSD cases in non-deployed group (n=6)Non-response bias assessed for demographics; significant differences between responders and non-responders on age, service branch, rank, sexOverall risk of bias: high
PTSD GWV vs non-deployed Adj OR = 4.74 (95% CI 2.05-10.94) adjusted for age, sex, income, education
Of participants, 86% of GWV and 87% of non-deployed group were male
Males and females included in the analysis
Gray 2002(63)
Cross-sectional postal survey; 1997 to 1999
Sample of all US Gulf War-era Seabees (members of US Naval Mobile Construction Battalions; Gulf War Seabees (n = 3,831) Seabees deployed elsewhere (DE) (n = 4,933) and non-deployed Seabees (ND) (n = 3,104)
Self-reported physician-diagnosed PTSD with onset after 1991
PTSD 3.08 0.71 (ND) 0.61 (DE)
Response rates: 68.6% overall (17,559 contacted.Non-response bias assessed in study: significant differences between responders and non-responders on select demographics and on health outcomesOverall risk of bias: high
PTSD GW Seabees vs ND Adjusted OR 4.23 (95% CI 2.59-6.92)PTSD Seabees vs Deployed elsewhere Adjusted OR = 4.27 (95% CI 2.79-6.52)
Adjusted for age, gender, active duty/reserve, status, race/ethnicity, current smoking and alcoholOf participants, 95.5% were male. Males and females included in the analysis
McCauley 2002(64)
Cross-sectional telephone survey; 1998 to 1999
Random sample of three groups of US Army or National Guard veterans living in 5 US states, and
Self-reported physician diagnosed PTSD since the Gulf War
PTSD 7.0 (KHAM) 0.8 (ND)8.7 (non-KHAM)
Response rates: 71% KHAM GWV;
Participation rates: 78% PTSD deployed v non-deployed adj OR = 14.9 (95% CI 5.6-60.9)
58
First author
Study design and period
Sample Case definition and measure
Outcome GWV prevalence (%)
Comp group prevalence (%)
Comments and risk of bias assessment
1.) Serving within 50k of Khamisiyah Iraqi munitions site (KHAM GWV n=653), 2) Non-KHAM deployed (GWV other deployed n=610) and 3) Non-deployed (n=516)
PTSD KHAM v non-KHAM adj OR = 1.0 (95% CI 0.6-1.7)Adjusted for age, gender, race, region of residenceOf participants, KHAM 96.2% males, non-KHAM 93.9% males, non-deployed 89.3% malesMales and females included in the analysis
KHAM (838 eligible), 75% non-KHAM (809 eligible), 38% non-deployed (741 eligible).Non-response bias was assessed in the study for demographics, health outcomes; significant differences between responders and non-responders. Sample was not representative of troops serving in the Gulf WarOverall risk of bias: high
Kang 2003(83)
Cohort postal/telephone survey as Phase 1 of the longitudinal National Health Survey of Gulf War era veterans and their Families; 1995-1996
Stratified random sample of US GV (n=11,441) and non-deployed (n=9476). Stratified by gender (females, n=3000), unit component (active, research and national guard) (NG=4000, reservists=5000) and branch (Army, Navy, Air force and Marine Corps)
PTSD Checklist (PCL) to assess current prevalence of PTSD, cut-off score ≥ 50
PTSD (cohort) 12.1 4.3 Participation rate: 70% overall (76.3% GWV vs 63.2% non-GWV) eligible in-state and contactable.Non-response bias assessed for demographics; significant differences between responders and non-responders on age, race, marital status and rankOverall risk of bias: low
PTSD (population)
10.1 4.2
PTSD in GWV vs non deployed adj OR = 3.1 (95% CI 2.7-3.4) adj for gender, age, marital, rank and unit component.Population rate accounted for complex sample survey design unadjusted OR 2.6 (95% CI 2.2-3.0)
GWV 81.4%, non-deployed 78.1% malesMales and females included in the analysis
Ikin 2004(1)
Cross-sectional in-person structured diagnostic interview admin-istered by trained psychologists; 2000 to 2002
All Australian GWV (n = 1381) and random sample of Navy, Army and Air force ND active duty personnel (n = 1377) matched by age, gender and service type
CIDI using DSM-IV criteria to assess prevalence PTSD first present post-Gulf War
PTSD 5.4 1.4 Participation rates: 81% GWV (1808 eligible); 57% non-GWV (2796 eligible).Non-response bias assessed in study: yes, for demographics, health outcomesSignificant differences between respondents and non-respondents: yes, on demographics; no on health
PTSD adj OR = 3.9 (95% CI 2.3-6.5) adjusted for service type, rank, age, education, marital status
Males only included in the analysis
59
First author
Study design and period
Sample Case definition and measure
Outcome GWV prevalence (%)
Comp group prevalence (%)
Comments and risk of bias assessmentoutcomesPsychological interview completion rates: 97% GWV; 89% non-deployedOverall risk of bias: low
Black 2004(17)
Cross-sectional structured telephone interview; 1995-1996
Random sample selected from US Iowa GWV; regular military (Reg) (n = 985), GW National Guard (NG)/Reserve (Res) (n = 911), non-GW regular military (n = 968), and non-GW NG/Res (n = 831)
PTSD Checklist (PCL) to assess current (past month); cut-off score of ≥ 50
PTSD (reg) 1.9 0.7 Participation rate: 91%, 76% (of eligible participants n=4886)Non response bias was not assessed. Limited to Iowans, limited power to detected differences between racial groupsSignificant differences between responders and non responders: not assessed.Limitations include Recall bias, sensitised by media/medical professionalsOverall risk of bias: Low
PTSD (NG/Res
2.0 1.1
PTSD (total) 1.9 0.8
Prevalences were weighted to account for complex study design.PTSD OR in Gulf War veterans vs non-deployedReg: OR 2.6 (95% 1.19-6.2)NG/Res OR 1.9 (1.0-3.5)Total: OR 2.5 (95% CI 1.2-5.0)adjusted for age, gender, race, rank, branch, military status, and prior mental health conditionProportion of females in sample not reported.Participation in combat was related to PTSD OR =2.1 (95% CI 1.7-4.2)
Fiedler 2006(66)
Cross-sectional telephone structured diagnostic interview administered by trained interviewers; 2000 to 2001
Random sample (n = 967) of all US GWV and non-deployed personnel (era veterans) (n = 784) including Army, Navy, Air Force, Marines and Coast Guard
12-month version of CIDI for the DSM-IV criteria to assess PTSD
PTSD (total) 3.4 0.9 Response rates: 59% GWV (1651 eligible); 51% non-deployed (1552 eligible).Non-response bias assessed in study: yes, for demographicsSignificant differences between respondents and non-respondents: yesAdj OR presented for anxiety disorders but not for PTSD separately in paperOverall risk of bias: low
PTSD (males) 3.4 0.7
PTSD (females)
4.0 2.2
GWV 92% males. Era veterans 88% males
Toomey 2007
Cross-sectional in-person
Stratified random subsample of previous
PTSD diagnosis using CAPS, a
PTSD diagnosis
6.2 1.1 Participation and Psychological interview completion rates:
60
First author
Study design and period
Sample Case definition and measure
Outcome GWV prevalence (%)
Comp group prevalence (%)
Comments and risk of bias assessment
(67) structured diagnostic interview, administered by trained interviewers; 1998 to 2001
study (68)
US GWV (n = 1061) and non-deployed (n = 1128).
structured interview, to DSM-IV criteria, for Gulf –era onset (Jan 1991 to July 1993) PTSD
Current PTSD symptom severity assessed using a 17-item PTSD checklist rated on scale of 1-5
53% GWV (1996 eligible); 39% non-GWV (2883 eligible)Non-response bias assessed in study: yes, for demographics, health outcomesSignificant differences between respondents and non-respondents: yes, on demographics; no on health outcomesOverall risk of bias: low
PTSD Checklist caseness (≥50)
9.8 2.7
PTSD Checklist mean score (SE)
27.5 (0.6) 22.8 (0.4)
PTSD diagnosis adj OR = 5.78 (95% CI 2.62-12.74) adjusted for age, gender, ethnicity, duty type (active v reserve/guard), service branch and rank.Current PTSD checklist caseness adj OR 3.84 (95% CI 2.16-6.80) adjusted for age, gender, education, ethnicity, duty type, service branch and rankPTSD checklist mean score was significantly greater in GWV compared to non-deployed personnelMales and females included in the analysisDeployed and non-deployed groups 78% males
Murphy 2006(84)
Cross-sectional postal survey. A study to develop and evaluate screening questionnaires in the military, without framing of Gulf War deployment context; 2001
Stratified random sample of UK Armed forces services (Army, Navy, RAF) by service and size of unit, 45 individuals randomly selected from each unit. GWV status was determined by record linkage. GWV (n = 308), non-GWV (n = 1339)
PTSD Checklist (PCL-C), cut-off ≥50 on full 17 item PCL-C or ≥40 on abridged 14 item PCL-C version. Full and abridged survey questionnaires tested in both study groups
PTSD 1.0 1.9 Participation rate: overall approximately 57% (out of 2873 in original sampling frame were linked on GWV status)Non-response bias not assessedComparison between groups displayed differences on service with larger proportion of Army, fewer Navy and RAFOverall risk of bias: low
PTSD Adjusted OR = 0.42 (95% CI 0.12-1.42) adjusted for length of questionnaire (where data from both questionnaires was available), age, gender, rank, service
GWV 99.4% males, non-GWV95.2% males. Males and females included in the analysis
Kang 2009(31)
Cross-sectional postal and telephone survey; 2004
Follow-up stratified random sample from previous study (68) of US GWV (Navy, Army, Air Force, Marines; n = 6111) and non-deployed Gulf Era personnel frequency
PCL-C to assess current symptom severity of PTSD in past 4 weeks; cut-off score of ≥ 50
PTSD 15.2 4.6 Response rates: overall 34%; 40% GWV (15,508 eligible); 27% Gulf Era (14,494 eligible).Non-response bias assessed for demographics, health outcomes; significant differences between
61
First author
Study design and period
Sample Case definition and measure
Outcome GWV prevalence (%)
Comp group prevalence (%)
Comments and risk of bias assessment
matched on gender, service branch and status (n = 3859)
responders and non-responders on demographics but not on health outcomesOverall risk of bias: lowAdjusted Risk Ratio (RR) = 2.98 (95% CI 2.54-3.50),
adjusted for age, gender, race, BMI, current smoking, rank, service branch, unit component (active duty/NG/guard/ reserve)Of participants, 79.9% GWV and 78.2% Gulf Era were males. Males and females included in the analysis
Smith 2009(86)
Baseline survey of US Millennium cohort Study; 2001-2003
75,156 baseline consenting participants of Millennium Cohort Study (enrolled July 2001 to June 2003) consisting of Army, Air Force, Navy/Coast Guard and Marine Corp.
Classified by deployment: none prior; 1991 GW only; 1991 GW and Bosnia/Kosovo/Southwest Asia; and Only Bosnia/Kosovo/Southwest Asia
PCL-C assessed current symptoms: cut-off ≥50 and ≥1 intrusion, ≥3 avoidance and ≥2 hyperarousal symptoms.Self-report measure assessed PTSD diagnosis: ever being told by their doctor or other health professional that they have PTSD (no, yes); if yes, what year did problem begin?
PTSD diagnosis, without current symptoms
18.3 59.4 Response rate: 71% (participation in follow up survey); 36% (to initial response to the participation)
No differences in responder health with respect to medical encounters
Differential response rates that are influenced by PTSD results
Potential response bias to continued follow-up is ongoing
Use of survey instrument as a surrogate for PTSD diagnosis and can only measure PTSD in a populationOverall risk of bias: low
PTSD symptoms without diagnosis
9.7 64.0
PTSD diagnosis with current symptoms
23.6 58.9
Comparisons in GW only and None prior: Weighted percent (above)PTSD diagnosis without current symptoms adj OR = 1.46 (95% CI 0.90-2.36)PTSD symptoms without diagnosis adj OR = 0.79 (95% CI 0.55-1.13)PTSD diagnosis with current symptoms adj OR = 1.33 (95% CI 0.59-2.99)
Weighted percent and OR adjusted for gender, birth year, education, marital status, race/ethnicity, military rank and component
Millennium Cohort participants 73.2% males. Males and females included in the analysis
62
4.4.2 Key findings
Eighteen studies were eligible for inclusion in the meta-analysis of PTSD in Gulf War
veterans. Pooled analysis based on random-effects model indicated that PTSD was over
three times more likely in Gulf War veterans compared with non-deployed military personnel.
There was moderate statistical heterogeneity between studies. Sources of heterogeneity
were assessed by subgroup analyses according to the outcome measure used, and
stratification by type of outcome measure used, duty status, risk of bias and adjustment of
OR did not reveal important differences between subgroups, but the summary OR of PTSD
in Gulf War veterans was higher in studies assessed as overall high risk of bias compared
with those with low risk of bias.
63
4.5 PTSD in Afghanistan/Iraq War veterans compared to non-deployed military personnel
4.5.1 Results
Eighteen studies met the eligibility criteria (3, 69-71, 73-75, 77, 79, 80, 89-96). Of these,
three studies (Fear, 2010, Hotopf, 2006, Jones, 2013) (3, 90, 92) had completely
overlapping samples so we only included the data for Fear et al. (3) in the meta-analysis.
Four studies (Fear 2010, Shen 2010, Smith 2008, Vasterling 2006) (3, 80, 94, 95) reported
separate ORs for different services (e.g. for regular and reservist veterans, or for Army, Navy
and Air Force veterans). In the main analysis we combined these within-study ORs so that
only one effect estimate per study contributed to the meta-analysis.
Main analysis
Based on a random-effects meta-analysis, there was an increased odds of PTSD in
deployed veterans compared with non-deployed personnel (OR 2.12, 95% CI 1.65 to 2.72)
(Figure 16). However, there was a very high amount of statistical heterogeneity (I2 = 97%),
so the meta-analytic effect should be interpreted with caution in terms of the actual level of
the increased odds.
64
Figure 16 Random-effects meta-analysis of PTSD in Afghanistan/Iraq War veterans and non-deployed military personnel
Subgroup and sensitivity analyses
Sources of heterogeneity were assessed by subgroup analyses according to the outcome
measure used (structured interview versus screening instrument), type of service (regular
versus reservist), risk of bias and adjustment of ORs.
Stratification by type of outcome measure did not reveal important differences between
subgroups (Figure 17). Results of the meta-regression suggested no statistically significant
association between type of outcome measure and magnitude of OR (P = 0.82).
Figure 17 Random-effects meta-analysis of PTSD in Afghanistan/Iraq War veterans and non-deployed military personnel, subgrouped by type of outcome measure
65
Stratification by regular versus reservist service did not reveal important differences between
subgroups (Figure 18). Results of the meta-regression suggested no statistically significant
association between type of service and magnitude of OR (P = 0.88).
Figure 18 Random-effects meta-analysis of PTSD in Afghanistan/Iraq War veterans and non-deployed military personnel, subgrouped by type of service
Stratification by risk of bias did not reveal important differences between subgroups (Figure
19). Results of the meta-regression suggested no statistically significant association
between risk of bias and magnitude of OR (P = 0.62).
66
Figure 19 Random-effects meta-analysis of PTSD in Afghanistan/Iraq War veterans and non-deployed military personnel, subgrouped by risk of bias
Stratification by adjustment of ORs did not reveal important differences between subgroups
(Figure 20). Results of the meta-regression suggested no statistically significant association
between adjustment of OR and magnitude of OR (P = 0.84).
67
Figure 20 Random-effects meta-analysis of PTSD in Afghanistan/Iraq War veterans and non-deployed military personnel, subgrouped by adjustment of odds ratios
Sensitivity analyses indicated that the overall OR did not vary after excluding any individual
study, and the statistical significance and heterogeneity did not change (Table 6).
68
Table 6 Sensitivity analyses excluding each study one by one for studies of PTSD in Afghanistan/Iraq War veterans compared to non-deployed military personnel
Excluded study Pooled ES LCI 95% HCI 95% Cochran Q Chi2 P-value I2 (%)
Barlas 2011 2.18 1.70 2.79 450.90 0.00 97
Bleir 2011 2.12 1.63 2.74 512.02 0.00 97
Bray 2006 2.21 1.72 2.84 466.71 0.00 97
Bray 2009 2.17 1.69 2.78 452.65 0.00 97
Fear 2010 2.20 1.71 2.83 480.55 0.00 97
Hoge 2004 2.16 1.67 2.77 456.89 0.00 97
Hoge 2006 1.99 1.56 2.53 266.92 0.00 95
Hourani 2007 2.11 1.63 2.73 517.15 0.00 97
Kline 2010 2.04 1.57 2.64 520.47 0.00 97
Peterson 2010 2.03 1.57 2.61 518.96 0.00 97
Shen 2009 2.02 1.47 2.79 519.22 0.00 97
Shen 2010 2.11 1.62 2.74 492.42 0.00 97
Smith 2008 2.04 1.55 2.67 520.64 0.00 97
Vanderploeg 2012 2.23 1.73 2.86 506.99 0.00 97
Vasterling 2006 2.21 1.71 2.84 495.76 0.00 97
Wittchen 2012 2.10 1.63 2.71 520.26 0.00 97
The funnel plot was asymmetrical (Figure 21), and the Egger test was statistically significant (P =
0.003), which suggests that small studies had systematically different ORs to larger studies. (Note:
FE MA, Fixed effects meta-analysis)
Figure 21 Funnel plot for the random-effects meta-analysis of PTSD in Afghanistan/Iraq War veterans and non-deployed military personnel
69
Comparison of summary odds ratio of PTSD in Gulf War veterans vs Afghanistan/Iraq War veterans
The 95% CI for the OR of PTSD in Gulf War veterans (OR 3.39, 95% CI 2.79 to 4.13) and
Afghanistan/Iraq War veterans (OR 2.12, 95% CI 1.65 to 2.72) did not overlap. The P-value for the
test for equality of the summary OR of PTSD in Gulf War veterans’ meta-analysis and summary
OR of PTSD in Afghanistan/Iraq War veterans’ meta-analysis was 0.004, indicating that the
summary OR of PTSD in Gulf War veterans was statistically significantly higher.
70
Table 7 Characteristics of eligible studies comparing prevalence of PTSD in Afghanistan/Iraq War veterans and non-deployed military personnel
First-named author
Study design, study period
Sample PTSD case definn, measure
Outcome Prevalence (%)
Comp group prevalence (%)
Comments and risk of bias assessment
Shen 2009(93)
Retrospective database analysis; data on US sailors deployed overseas 2002-2006, including Defense Manpower Data Center and Post-Deployment Health Assessment (PDHA) survey data
Sailors deployed in a variety of overseas locations (n=112,720), including Iraq, Afghanistan, other ground deployments, with deployed on a ship as reference category
4-item PDHA survey to assess positive screening for PTSD, cut-off ≥2 items
PTSD Iraq 10.0 - Participation rate: High PDHA completion for Afghanistan/Iraq deployed but sailors on routine ship operation exempted. Iraq/ Afghanistan deployed more likely male and middle ranked sailors vs those on ships. PDHA administered soon after deployment, may not fully capture PTSD not manifested. Possible underreporting because stigmaOverall risk of bias: high
PTSD Afghan 5.0 -
PTSD Other - 3.0
PTSD Ship - 3.0
Models presented for effect of deployment location on probability of screening positive for PTSD but no ORIn participants, overall 88% males, Iraq 92% males, Afghanistan 93% males Other 86% males, Ship 88% males. Males and females included in the analysis
Wittchen, 2012(96)
Cross sectional study; 2010-2011
Stratified random sample German soldiers deployed in Afghanistan (n=1483), and comparison not deployed overseas (n=889)
Computer assisted Munich CIDI-M (military) interview by psychologist to assess PTSD to DSM-IV criteria, 12 month prevalence, 12 month incidence, lifetime prevalence
12 month diagnosis
2.9 1.2 Participation rate: 92.8% deployed, 95.4% non-deployed.Non response bias not assessed, but study group assessed representative of German armed forces deployed to AfghanistanOverall risk of bias: low
12 month incidence
0.9 0.2
Lifetime diagnosis
4.6 2.7
12 month diagnosis OR = 2.5 (95% CI 1.1-5.6)12 m incidence OR = 4.2 (95% CI 0.7-24.5)Lifetime prevalence OR = 1.7 (95% CI 0.96-3.1)Prevalences were weightedDeployed 94.8%, non-deployed 95.5% males. Males and females included in analysis
Kline, 2010(75)
Cross sectional anonymous self-administered survey pre-deployment to Iraq: 2007-08
New Jersey (NJ) National Guard (NG) members (n=2543): ≥1 prior OEF/OIF deployment since 2001 (n=625) and No prior OEF/OIF deployments (n=1910)
PTSD Checklist (PCL): cut-off ≥50 (more restrictive) and symptom cluster method based on DSM-IV (less restrictive)
PTSD (PCL) 14.0 4.2 Participation rate: 95% overall.Non response bias not assessed due to survey anonymity. NJ NG similar to national NG sample on some demographics and PTSD, but included greater proportion of Hispanics, fewer non-Hispanic whites, though few differences in current data on raceOverall risk of bias: low
PTSD (DSM –IV)
21.1 9.0
PTSD (PCL) OR = 3.69 (95% CI 2.59-5.24)PTSD (DSM – IV) OR =2.70 (95% CI 2.05-3.55) adjusted for age, sex, race/ethnicity, education, income, marital status and military deployment other than in OEF or OIF
Of participants, deployed 85.5% males, non-deployed 89.0% males. Males and females included in the analysis
71
First-named author
Study design, study period
Sample PTSD case definn, measure
Outcome Prevalence (%)
Comp group prevalence (%)
Comments and risk of bias assessment
Vanderploeg; 2012(79)
Cross sectional anonymous online survey; 2009-10
Current active members of Florida NG deployed to OEF/OIF (n=1443) and not deployed (n=1655)
PCL-C, cut-off score ≥50 and reported very difficult or greater to function at work, home or get along with others
Probable PTSD
6.9 1.9 Participation rate: 41.3%Non response bias not assessed due to survey anonymity No information on Florida NG cohort available.Overall risk of bias: high
PTSD OR = 0.40 (95% CI 0.13-1.23) adjusted for other covariates, demographic, pre-deployment factorsOf participants, deployed 87.4% males, non-deployed 79.2% males. Males and females included in analysis.
Shen, 2010(94)
Database analysis on several combined sources from US Defense Manpower Data Center and TRICARE, to obtain active duty personnel, demographic service, mental and related health information
Active duty enlisted personnel serving b/n 2001 and 2006 (n=678,227); deployed to Afghanistan/Iraq, not deployed to Afghanistan/Iraq, deployed other OEF/OIF missions (eg Kuwait, Qatar)
ICD-9 coded clinical diagnosis of PTSD any time 2001 to 2006
PTSDArmyMarinesNavyAir Force
4.413.516.461.34
0.630.520.830.62
Sample 25% and representative of US Armed Forces active duty population, Captured diagnoses from those still in service, may have missed severe cases diagnosed within VA system, though this not likely to be biased by deploymentOverall risk of bias: low
PTSD Deployed vs not deployed Iraq/ Afghanistan based last deployment locationArmy adj OR 3.96, Marines OR 4.57, Navy adj OR 9.06, Air Force adj OR 1.25
PTSD Deployed other OEF/OIF missions vs not deployed Afghanistan/IraqArmy adj OR 3.97, Marines adj OR 3.51, Navy adj OR 0.54, Air Force adj OR 0.36adj for service, demographic characteristicsPersonnel included 84.1% (Air Force) to 96.3% (Marines) males. Males and females included in analysis
Bleier 2011(89)
Analysis of cross-sectional surveys of Australian personnel deployed to Solomon Islands, East Timor or Bougainville and military comparison groups; 2007-2008
Personnel who deployed at least twice; reported they deployed at least once to Iraq/ Afghanistan and East Timor (n=771) and never deployed (n=573)
PCL-C; cut-off ≥30 PTSD 28.0 20.0 Participation rate overall for surveys: 44%Data analyses limited to those with adequate complete deployment data. Non-response bias assessed: ex-serving and enlisted people under-represented in responders.PCL-C cut-off of ≥30 considered of clinical relevance, top quartile for combined dataset, because of relevance of subsyndromal PTSDOverall risk of bias: low
PTSD deployed at least twice (including Iraq/ Afghanistan and East Timor) vs never deployed adj OR 2.1 (95% CI 1.5-2.8) adj for age, sex, rank, current serving status
Of participants, 92% at least twice deployed, 84% never deployed males. Males and females included in analyses
Peterson, Cross-sectional Active duty US Air Post-Development PTSD 4.1 0.7 Previous reports showed high
72
First-named author
Study design, study period
Sample PTSD case definn, measure
Outcome Prevalence (%)
Comp group prevalence (%)
Comments and risk of bias assessment
2010(77)
survey post deployment of US personnel deployed to Iraq or Qatar based on de-identified data from US DoD PDHA surveys; 2005-2007
Force personnel deployed to Iraq (combat zone) (n=4,408) or Qatar ((non combat zone, ‘non deployed’ comparison group) (n=959) who completed a PDHA
Health Assessment (a 4-item measure to screen for PTSD in primary care, PC-PTSD Screen), score ≥2 positive screen for PTSD, score ≥3 higher threshold risk
positive screen
completion rates for Health Assessments but response rate not reported/applicable. Demographic characteristics of groups similar in gender, age, marital, and military grade, but average time in service unavailable. Deployment length = 4m. Generalisability for other services could be limitedOverall risk of bias: high
PTSD higher threshold risk
2.1 0.4
PTSD positive screen OR = 5.86 (95% CI 2.66-13.62).PTSD higher threshold risk OR = 5.00 (95% CI 1.78-16.12).Upper limits of 95% CI noted not accurate by Cornfield method. Adjustment factors not stated.Deployed to Iraq 85.4% males, deployed to Qatar 83.9% males. Males and females included in the analysis.
Hourani 2007 (91)
Cross-sectional, self-report questionnaire survey; 2006
Sample of all eligible US reserve and NG military personnel comprising army, navy, marine, air force reserve and coast guard. Served in OEF/OIF (n = 5325), non-deployed (n = 4886)
PCL-C past 30 days, cut off ≥50 considered need for further evaluation of PTSDDeployment status based on served in OEF or OIF, did not serve in theatres
PTSD 10.5 (SE 1.2) 4.9 (SE 0.7) Overall response rate: 51.8%. Nonresponse adjustment to help compensate for potential bias of nonsurveyed persons. Survey targeted to US reserve and NG personnel rather than military overall. Responses anonymous, so likely minimised non reporting biasOverall risk of bias: low
Adjusted prevalences (SE). Adjusted estimates were standardised to correct for differences in the demographic distributions between the theater of operations groups. Main effects of Reserve component, gender, age group, enlisted/officer status, marital status, education, and race/ethnicity were used in this standardisation process
Bray 2006 (69)
Cross sectional self-report anonymous DoD survey of Health Related Behaviour Among Active Duty Military Personnel (HRBS); 2006
Sample all eligible active duty US military personnel. n=16,146 (3639 Army, 4627 Navy, 3356 Marine, 4524 Air Force). 40,436 US active duty service members were sampled with 28,546 completing the surveys (5927 Army, 6637 Navy, 5117 Marine
PCL-C past 30 days, cut off ≥50 considered need for further evaluation of PTSDDeployment status based on served in OEF or OIF, did not serve in theatres
PTSD 7.1 (0.6) 6.1 (0.6) Overall response rate: 51.8%Participants were selected to represent men and women in all pay grades of the active force worldwide. Data were weighted to represent all active duty personnel in the analysesOverall risk of bias: low
Adjusted prevalences (SE). Adjusted estimates were standardised to correct for differences in the demographic distributions between the theater of operations groups. Gender, age group, enlisted/officer status, marital status, education, and race/ethnicity were used in this standardisation process
73
First-named author
Study design, study period
Sample PTSD case definn, measure
Outcome Prevalence (%)
Comp group prevalence (%)
Comments and risk of bias assessment
Crops, 7009 Air Force and 3856 Coast Guard)
Bray 2009 (70)
Cross sectional self-report anonymous Health Related Behaviour Among Active Duty Military Personnel questionnaire survey; 2008 (10th data point since 1980, included Coast Guard for first time)
Sample of all eligible active duty US military personnel (excluded recruits, academy students, AWOL or incarcerated personnel). n=28,546; (5927 Army, 6637 Navy, 5117 Marine, 7009 Air Force, 3856 Coast Guard)
PCL-C past 30 days, cut off ≥50 considered need for further evaluation of PTSDDeployment status based on combat deployed and served in OIF or OEF or Not combat deployed since Sept 11, 2001
PTSDAllArmyNavyMarinesAir ForceDoD ServicesCoast Guard
12.4 (SE 0.8)16.4 (SE 1.5)8.9 (SE 0.6)16.5 (SE 2.3)5.5 (SE 0.6)12.4 (SE 0.8)
10.7 (SE 2.9)
8.2 (SE 0.4)9.0 (SE 1.0)8.2 (SE 0.7)13.0 (SE 1.2)5.3 (SE 0.4)8.4 (SE 0.4)
5.9 (SE 0.5)
Overall response rate: 71.6%Participants were selected to represent men and women in all pay grades of the active force worldwide. Data were weighted to represent all active duty personnel in the analysesOverall risk of bias: low
Barlas 2011 (71)
Anonymous cross-sectional DoD web-based survey of Health Related Behaviour Among Active Duty Military Personnel (HRBS), 2011
Stratified random sample all 154,011 eligible members US Army, Navy, Marine Corps, Air Force, and 14,653 Coast Guard non-deployed and on active duty at time of study. 34,416 DoD respondents (Army 6,932, Navy 7,571, Marine Corps 8,339, Air Force 11,574); 5,461 USCG
4-items about feeling very upset when reminded of a stressful experience, emotional numbness/lack of loving feelings for close persons, difficulty concentrating, feeling jumpy/ easily startled, on a 1-5 Likert scale. Total score ≥4 = “high PTS” level past 30 days. Deployment status -combat deployed and served in OIF, OEF or New Dawn, Not combat deployed since Sept 11, 2001
PTSDAllArmyNavyMarinesAir ForceCoast Guard
5.9 (SE 0.3)8.6 (SE 0.6)3.6 (SE 0.5)6.7 (SE 0.6)2.3 (SE 0.2)3.5 (SE 1.0)
4.0 (SE 0.3)4.5 (SE 0.7)4.7 (SE 0.7)7.4 (SE 0.7)1.6 (SE 0.2)1.9 (SE 0.3)
Response rate: DoD 22%, Coast Guard 37%Survey designed to be a statistically-valid selection of a representative sample of service members. Data were weighted to represent all active duty personnel in the analysesOverall risk of bias: low
Hotopf 2006 Cohort study Stratified random PCL-C cut-off ≥50, PTSD (PCL-C)
4.0 3.7 Participation rates: 62.3% deployed
74
First-named author
Study design, study period
Sample PTSD case definn, measure
Outcome Prevalence (%)
Comp group prevalence (%)
Comments and risk of bias assessment
(90) self-report questionnaire; 2004-06
sample of UK armed forces personnel deployed to Iraq Jan to April 2003 (Op TELIC 1) (n=4722, 3936 regulars, 786 reservists) and UK armed forces personnel at the time but not deployed to TELIC 1 (Era) (n=5550, 4750 regular, 800 reservists).
and also a definition requiring the participant scored moderate or above on one of the re-experiencing symptoms, three avoidance symptoms, andtwo hyperarousal symptoms
Regular 56.3% non-deployed samples
Non-response bias assessed via intensive follow up study, late responder analysis, sensitivity analysis, healthy warrior effect considered – unlikely important difference missedOverall risk of bias: low
Reservists 6.0 2.8PTSD (alternate definition)
4.0 3.0
PTSD OP Telic vs Era Regulars Adj OR 1.17 (95% CI 0.92–1.48)Reservists Adj OR 6.95 (95% CI 0.89–54.2)Interaction for deployment by regular/reservist status for PTSD (p=0.02)PTSD OP Telic vs Era Regulars adj OR 1.00 (0.79–1.28) after reassigning Era participants to Iraq War group if they had served in subsequent TELIC 1 deploymentPTSD (alternative definition) OR 1.21 (95% CI 0.96–1.53)OR adj for age, sex, rank, educational and marital status, service branch, and fitness to deployTELIC 1 sample 92% males, Era 90% males. Males and females included in analyses
Hoge et al 2004 (74)
Anonymous cross-sectional survey phase of a longitudinal study; 2003
Samples from an Army combat infantry brigade pre deployment to Iraq (n=2530); an infantry brigade of same division 6m post deployment to Afghanistan (n = 1962); an Army infantry brigade 8m post Iraq deployment (n = 894); and Marine Corp units 6m post Iraq deployment (n = 815)
Questionnaires administered 3-4m after return to US. PCL. Results scored positive if ≥1 intrusion, 3 avoidance, 2 hyperarousal symptoms categorised as moderate level. For the strict definition to be met, score had to be ≥50
PTSD (broad)
18.0 (Army post Iraq)11.5 (Army post Afghanistan)19.9 (Marines post Iraq)
9.4 (Army pre deployment Iraq)
Response rate: (defined as completion of any part of survey) of 98% among the 58% of combined samples available to attend study briefings
Demographic characteristics of sample similar to the general, deployed, active duty infantry populationOverall risk of bias: low
Adj OR Army post Iraq 2.13 (95% CI 1.71–2.66)Adj OR Army post Afghanistan 1.25 (95% CI 1.03–1.52)Adj OR Marines post Iraq 2.40 (95% CI 1.92–2.99)All vs Army pre deployment to Iraq, adj for age, rank, education, marital, race/ethnicityPTSD (strict)
12.9 (Army post Iraq)6.2 (Army post Afghanistan)12.2 (Marines post Iraq)
5.0 (Army pre deployment Iraq)
Adj OR Army post Iraq 2.84 (95% CI 2.17–3.72)Adj OR Army post Afghanistan 1.26 (95% CI 0.97–1.64)Adj OR Marines post Iraq 2.66 (95% CI 2.01–3.51)All vs Army pre deployment to Iraq, adj for age, rank, education, marital, race/ethnicity
75
First-named author
Study design, study period
Sample PTSD case definn, measure
Outcome Prevalence (%)
Comp group prevalence (%)
Comments and risk of bias assessment
Of participants, males ≥98% across services
Hoge et al 2006 (73)
Analysis of Defence Medical Surveillance System (DMSS) database on all US Army soldiers and Marines who completed a Post Deployment Health Assessment; May 2003 to April 2004
Army soldiers and Marines post deployment to Iraq (n=222,620), Afghanistan (n=16,318), and other locations, e.g. Bosnia, Kosovo (n= 64,967) who completed the routine PDHA
4-item PTSD screen (Primary care-PTSD screen or “PCPTSD”). 4 questions cover key domains of PTSD: re-experiencing trauma,numbing, avoidance, and hyperarousal. Cut-off ≥2/4 items at risk of PTSD. Deployment locations - OEF, OIF or other (Bosnia, Turkey, Uzbekistan, Kosovo, on a ship, or other)
PTSD OIF 9.8%OEF 4.7%
2.1 % Demographics of study population and study findings using survey and electronic versions of the PDHA were similar so combined. 18% of study population did not have a PDHA were similar to those who had a PDHA by deployment location and demographics but were somewhat more likely to be active duty Marines. Some demographic differences between deployment groupsOverall risk of bias: low
Adj OR OIF vs other 5.51 (95% CI 5.20-5.83)Adj OR OEF vs other 2.52 (95% CI 2.30-2.76)OR adjusted for sex, age, marital, service branch, component, grade
Of participants, OIF males (89.4)%, OEF males (91.3), other (88.6)
Vasterling, 2010(97)
Prospective pre and post deployment cohort survey
US Army deployed soldiers (n=774, regular active duty n=670, NG n=104), non-deployed (regular active duty n=309). Pre and post deployment assessment
PCL-C according to DSM-IV-TR, cut-off score ≥50
PTSD post deploymentRegularNG
12.014.0
11.0
Participation rate: 94% at time 1, 73% at time 2
Participants at Time 1 generally reflected US Army population but females and officers underrepresented. Sample not population based, only included one service branch. NG sample does not generalise to broader reservistsOverall risk of bias: low
Of participants, deployed 92.9% males and non-deployed 90.3% males. Males and females included in the analysis.
No OR included for PCL caseness, and therefore no adjustment for possible confounding factors. Study’s main outcome of interest was pre to post deployment change in PTSD severity by PCL-C score.
Smith, 2008(95)
Prospective cohort analysis at follow up in 2004-06; baseline 2001-03
Millennium cohort study participants – active duty and Reserve/NG personnel; (n=50,184 for analysis) (n=50128) deployed (n=11952) and non-deployed
PCL-C.Sensitive definition of symptoms of PTSD used DSM-IV criteria alone
Specific definition included DSM-IV criteria and cut-off ≥50
Persisting symptoms by sensitive criteria
47.9 (deployed with combat exposures)22.4(deployed without combat exposures)
45.9(Non deployed)
Participation rate in follow up: 71%
Analyses have previously suggested a representative sample of military personnel by demographic and health outcomes, shown minimal reporting biases, and on differences in methods of completing surveyOverall risk of bias: low
New onset symptoms by sensitive criteria
8.7(deployed with combat exposures)2.1(deployed without combat exposures)
3.0(Non deployed)
76
First-named author
Study design, study period
Sample PTSD case definn, measure
Outcome Prevalence (%)
Comp group prevalence (%)
Comments and risk of bias assessment
(n=38176) Self-reported having been told by their doctor or other health professional that they have PTSD (‘ever’ at baseline, ‘past 3 years’ at follow up)
Persisting symptoms by specific criteria
43.5(deployed with combat exposures)26.2(deployed without combat exposures)
47.6(Non deployed)
New onset symptoms by specific criteria
4.3 (overall)7.6(deployed with combat exposures)1.4(deployed without combat exposures)
2.3(Non deployed)
PTSD specific definition deployed with combat exposures vs not deployedArmy adj OR 3.59 (95% CI 3.08-4.17)Air Force adj OR 3.38 (95% CI 2.29-4.98)Navy/CG adj OR 2.48 (95% CI 1.48-4.14)Marine adj OR 2.78 (95% CI 1.52-5.07)Deployed without combat exposures vs not deployed Air Force adj OR 0.56 (95% CI 0.3-0.89) Other comparisons not significant.OR adjusted for baseline characteristics including sex, age, education, marital status, race/ethnicity, rank, service component, occupation, smoking, and problem alcohol drinking.Of participants, deployed 81.4% males and non-deployed 69.6% males. Males and females included in the analysis.
Fear, 2010(3)
Cohort study survey; 2007-09. Multiple phase study- 2004-06 sample was reassessed and 2 new samples added
Randomly sampled UK armed forces (n=9990, reg regulars, 8278, res, reservists, 1712). Included Iraq cohort (90) and personnel deployed Afghanistan (April 06-07), and joined UK armed forces since 2003;
PCL C, cut-off ≥50 PTSDI only
4.8 4.0 Participation rate: 56% (of eligible sample)Non-response bias assessed: no association between mental health at phase 1 in and responding phase 2 in Iraq cohort. Analyses considered response weightsOverall risk of bias: low
PTSDA only
3.4 4.0
PTSD Both I and A 2.7 4.0Regulars PTSD I/A
4.2 4.0Reservists PTSD I/A
5.0 1.8PTSD deployed I only vs not deployed I/A adj OR = 1.20 (95% CI 0.87-1.67)PTSD A only vs not deployed I/A adj OR = 0.93 (95% CI 0.54-1.59)PTSD Both I/A vs not deployed I/A adj OR = 0.92
77
First-named author
Study design, study period
Sample PTSD case definn, measure
Outcome Prevalence (%)
Comp group prevalence (%)
Comments and risk of bias assessment
deployed Iraq only (I) (n=4203), Afghanistan only (A) (n=1123), Iraq and Afghan-istan (I and A) (n=1389), Not deployed Afghanistan/Iraq (n=3255)
(95% CI 0.58-1.46)PTSD Regulars deployed I/A vs not deployed I/A adj OR = 1.13 (95% CI 0.82-1.54)PTSD Reservists deployed I/A vs not deployed I/A adj OR = 2.83 (95% CI 1.23-6.51)
OR adjusted for age, sex, marital status, educational status, rank, serving status, service.
Of participants, deployed to I only 90.6% males, A only 94.5% males, I and A 94.3% males, not deployed I/A 86.8% males. Males and females included in the analysis.
78
4.5.2 Key findings
Sixteen studies were eligible for inclusion in the meta-analysis of PTSD in Afghanistan/Iraq War
veterans. Pooled analysis based on a random-effects model indicated that PTSD was just over
twice as likely in Afghanistan/Iraq War veterans compared with non-deployed military personnel.
However there was high heterogeneity between studies and the level of the elevated odds of the
summary meta-analysis effect estimate should be interpreted with caution. Sources of
heterogeneity were assessed by subgroup analyses. Stratification according to the outcome
measure used (structured interview versus screening instrument), type of service (regular versus
reservist), risk of bias and adjustment of ORs did not reveal important differences between
subgroups and the high level of heterogeneity persisted. Gulf War veterans had a higher odds of
PTSD than the Afghanistan/Iraq War veterans relative to their non-deployed comparison groups.
79
4.6 Alcohol use and substance use disorders in Gulf War and Afghanistan/Iraq War veterans compared to non-deployed military personnel
4.6.1 Results
Of the 25 primary studies identified in relation to psychological disorders in Gulf War veterans
(Figure 1), nine studies (1, 7, 8, 29, 31, 64, 66, 67, 84) were included in the meta-analysis in
relation to alcohol use and substance use disorders in Gulf War veterans. Of the 24 primary
Afghanistan/Iraq War veteran studies identified, nine studies reported on alcohol/substance use
disorders and seven discrete studies were included in the meta-analysis. The characteristics of
these studies are reported in Table 8 and Table 9. Across the Gulf War studies, sample sizes
ranged from 308 to 6111 for Gulf War veterans and 482 to 3859 for the non-deployed personnel.
Across the Afghanistan/Iraq War veteran studies, other than the anonymous surveys or database
analyses, sample sizes ranged from 625 to 11,171 for Afghanistan/Iraq War veterans and 889 to
37,310 for non-deployed personnel. The percentage of female veterans in these studies was low.
It ranged from 0.6% to 22% for Gulf War veterans and 2.5% to 22% for the non-deployed military
comparison groups, and 3% to 21% for Afghanistan/Iraq War veterans and 4% to 55% for the non-
deployed military comparison groups.
The studies included Gulf War veteran cohorts from the US, UK, Australia and Canada, in all three
services (Navy, Army, and Air Force). The military comparison groups in all the studies were
described and defined as non-deployed, i.e. not deployed to the Gulf War or Afghanistan/Iraq War
during the period of operations, rather than other conflict/other deployed personnel e.g. Germany,
Bosnia. The Afghanistan/Iraq War veteran study populations were from the US, UK and Germany.
Some of the Afghanistan/Iraq War veteran studies considered associations between levels of
combat exposure and health outcomes within the deployed group, but the primary comparisons in
the meta-analysis are between the deployed and non-deployed group.
Of the Gulf War studies, three used structured diagnostic interviews to determine the caseness of
alcohol/substance use disorders (1, 66, 67) and four used screening tools (7, 8, 31, 84). The other
two studies (29, 64) used self-reported physician diagnosis. Of the seven studies for the outcome
of alcohol use disorders, three studies provided data for adjusted OR in the published article (1, 67,
84) and the study by Kang et al. (31) provided a Relative Risk, for which the unadjusted Relative
Risk was converted to an unadjusted OR with 95% CI. Other studies provided usable data in the
published article for calculation of an unadjusted OR (7, 8, 66). Of the three studies included in the
meta-analysis for the outcome of substance use disorders, two studies provided data for adjusted
80
OR (1, 67) and the remaining study provided usable data for calculation of an OR. Another two
studies (29, 64) provided data on alcohol or drug use disorders combined which was difficult to
separate; thus, they were not included in the meta-analysis of individual outcomes but were
included in the analysis of any substance use disorder outcome.
Of the Afghanistan/Iraq War veteran studies, one used a structured diagnostic interview to
determine alcohol/substance use disorder caseness (98), and the remaining eight studies used
screening tools. Hotopf et al. (90) provided informative data on UK veterans but the study
population formed part of the cohort of a later study of UK veterans (3) which was included in the
meta-analysis. Bray et al. (69, 70) provided on request more precise estimates of standard errors
of prevalence to enable us to calculate ORs.
Gulf War veteran studies - Alcohol use disorders
The overall OR using the random effects model (46, 47) for the seven included studies was 1.33
(95% CI: 1.22, 1.46) which indicated higher risk of alcohol use disorders in Gulf War veterans
compared to non-deployed military personnel. Overall heterogeneity for all studies represented by
I2 was 14% (Figure 22). Stratification by caseness indicated that the screening tool subgroup (OR
= 1.30, 95% CI: 1.14, 1.48) and the structured diagnostic interview subgroup (OR = 1.45, 95%:
1.17, 1.81) had comparatively similar ORs, and meta-regression indicated no statistically
significant association with caseness (p=0.452). However, heterogeneity was higher in the
subgroup of studies using screening tools (I2 = 46% versus I2 = 0%). The screening tool subgroup
contributed much greater weight in calculating the overall OR than the studies using diagnostic
interviews. There were no studies in the self-reported physician diagnosis subgroup for this
outcome.
Figure 22 Random-effects meta-analysis illustrating log-transformed odds ratios of alcohol use disorders in Gulf War veterans and non-deployed military personnel
81
The studies were further stratified based on adjusted versus unadjusted OR estimates, and the OR
in both subgroups were comparatively similar (adjusted subgroup OR = 1.38, 95% CI: 1.10, 1.74;
unadjusted subgroup OR = 1.31, 95% CI: 1.17, 1.49; results of meta-regression indicated no
statistically significant association with type of OR (p=0.815). Sensitivity analyses indicated that
the overall OR did not vary after excluding any individual study for this outcome, and the statistical
significance did not change. The funnel plot was visually inspected and it did not show an obvious
lack of symmetry, and the Egger test was not statistically significant (p=0.134).
Gulf War veteran studies – Substance use disorders
Three studies (1, 66, 67) reported a substance use disorder outcome separately and the overall
OR using random effects model for these studies was 2.13 (95% CI: 0.96, 4.72; I2 = 29%). The
result was not statistically significant. Sensitivity analysis indicated, after excluding the study by
Fiedler et al. (66), that the overall OR decreased to 1.80 (95% CI: 1.09, 2.98) with overall
heterogeneity reduced (I2 = 0%) and that this finding was statistically significant. The study by
Fiedler et al. (66) had only one subject in the Era veterans’ category for the substance use
disorders outcome, and this was the reason for excluding it in the sensitivity analyses. All studies
used structured diagnostic interviews to define caseness in substance use disorders. Stratification
of studies reporting a substance use disorder outcome in Gulf War veterans was not possible due
to the small number of studies.
Gulf War veteran studies - Any substance use disorder
Many included studies in this review reported dependence or abuse for more than one substance
(including alcohol) (1, 66, 67) and some reported OR for (combined) any substance use disorders
(29, 64), which prompted us to analyse these studies for the outcome of ‘any substance use
disorder’. The overall OR for the nine studies using random effects model for any substance use
disorder was 1.35 (95% CI: 1.25, 1.46) which was statistically significant and had low
heterogeneity of I2 = 0%, indicating a higher risk of any substance use disorders in Gulf War
veterans compared to non-deployed military personnel (Figure 23). Sensitivity analyses indicated
that the overall estimate did not change after excluding any individual study. The OR remained
statistically significant throughout these analyses. No asymmetry was detected in the funnel plot
and the Egger test was not statistically significant (p=0.863).
82
Figure 23 Random-effects meta-analysis illustrating log-transformed odds ratios of any substance use disorders in Gulf War veterans and non-deployed military personnel
Gulf War veteran studies - Risk of bias of included studies
All studies were assessed for overall risk of bias. Two (29, 64) of the nine studies received a high
overall risk of bias assessment. None of the seven studies included in the alcohol use disorder
outcome analysis were assessed as having a high overall risk of bias. The summary estimates for
any substance use disorders were further analysed based on studies having an overall low or high
risk of bias (42). The OR of any substance use disorder was 1.60 (95% CI: 0.96, 2.68) for the high
risk of bias subgroup and 1.34 (95% CI: 1.22, 1.47) for the low risk of bias subgroup. Meta-
regression indicated no statistically significant association with study risk of bias (p=0.516).
Afghanistan/Iraq War veteran studies - Alcohol use disorders
For the outcome, alcohol use disorders, the overall OR using the random effects model was 1.36
(95% CI: 1.11, 1.66) which indicated higher risk of alcohol use disorders in Afghanistan/Iraq War
veterans compared to non-deployed military personnel. Overall heterogeneity for all studies
represented by I2 was 77% (Figure 24). The studies were stratified according to duty status of
personnel (regular versus reservist) and the OR in the reservist subgroup was slightly higher
(regular subgroup OR = 1.32, 95% CI: 1.04, 1.68; reservist subgroup OR = 1.58, 95% CI: 1.31,
1.91) although results of meta-regression indicated no statistically significant association with duty
status of personnel (p=0.599). Stratification by adjusted versus unadjusted OR estimates also did
not identify important differences between subgroups (adjusted subgroup OR = 1.29, 95% CI: 1.02,
1.62; unadjusted subgroup OR = 1.66, 95% CI: 1.31, 2.09); results of meta-regression indicated no
statistically significant association with type of OR (p=0.348). Stratification by caseness was not
83
possible because all studies except one used a screening tool. Sensitivity analyses indicated that
the overall OR did not vary after excluding any individual study for this outcome, and the statistical
significance did not change. When meta-analysis was restricted to studies at low risk of bias, there
was no important change to the OR (OR = 1.45, 95% CI: 1.23, 1.72). The funnel plot was not
asymmetrical, and the Egger test was not statistically significant (p=0.338).
Figure 24 Random-effects meta-analysis illustrating log-transformed odds ratios of alcohol use disorders in Afghanistan/Iraq War veterans and non-deployed military personnel
Afghanistan/Iraq War veteran studies –Substance use disorders
For the outcome, substance use disorders, the overall OR using random effects model was 1.14
(95% CI: 1.04, 1.25), which indicated higher risk of substance use disorders in Afghanistan/Iraq
War veterans compared to non-deployed military personnel (Figure 25). There was no statistical
heterogeneity (I2 = 0%), so subgroup analyses were not necessary. Sensitivity analyses indicated
that the overall OR did not vary after excluding any individual study for this outcome, and the
statistical significance did not change. We could not perform a sensitivity analyses based on study
risk of bias as all studies were rated at low risk of bias. The funnel plot was symmetrical and the
Egger test was not statistically significant (p=0.476).
84
Figure 25 Random-effects meta-analysis illustrating log-transformed odds ratios of substance use disorders in Afghanistan/Iraq War veterans and non-deployed military personnel
Afghanistan/Iraq War veteran studies - Any substance use disorder
Shen 2012 (78) was the only study that measured an ‘any substance use’ outcome in
Afghanistan/Iraq War veterans, and combining these results with the alcohol use or substance use
disorder results of other studies produced a meta-analysis with very high heterogeneity (I2 = 98%),
which we were unable to explain with subgroup analyses (duty status of personnel (regular versus
reservist), adjusted versus unadjusted subgroup, structured diagnostic interview versus screening
tool for identifying caseness, sensitivity analyses, or risk of bias) (data not shown). Therefore, we
considered it more appropriate to report the results of Shen et al. (78) separately. Shen et al. (78)
found a higher risk of any substance use disorder in Army (OR 4.05, 95% CI: 3.82, 4.30), Marine,
(OR 4.36, 95% CI: 3.82, 4.97), Navy (OR 1.77, 95% CI: 1.45, 2.16) and Air Force (OR 1.76, 95%
CI: 1.56, 1.99) veterans of the Afghanistan/Iraq Wars compared to non-deployed military
personnel.
Comparison of summary odds ratio of alcohol use disorders and substance use disorders in Gulf War veterans versus Afghanistan/Iraq War veterans
Results of meta-regressions suggested that there was no statistically significant association
between theatre of war and the OR for alcohol use (p=0.862) and narrowly missed statistical
significance for the OR for substance use disorders (p=0.053).
85
Table 8 Characteristics of eligible studies comparing prevalence of alcohol/substance use disorders in Gulf War veterans and non-deployed military personnelAuthor Year (Reference)
Study design; study period
Sample Alcohol / substance use case defn measure
Outcome GWV prevalence
(%)
Comp group prevalence
(%)
Effect measure 95% Confidence Interval
Participation rates, Assessment of overall risk of bias
IOWA Persian Gulf Study Group 1997 (8)
Cross-sectional telephone interview; 1995-96
Random sample of US Iowa state Regular military (R) and National Guard/Reserve (NG/Res) GWV (n = 1896) and active duty or activated non-deployed personnel (n = 1799) stratified by service, age, sex, ethnicity and rank
CAGE screening questionnaire assessing symptoms of alcohol abuse
Symptoms of alcohol abuse (R)
17.0 12.2Prevalence difference in all subjects
2.4 0.4, 4.5Participation rates: 78% GWV (2421 eligible); 73% non-GWV (2465 eligible), 75.5% of males (4453 eligible), 77.4% of females (433 eligible) completed the interviews of all GWV and non-GWV.Non-response bias was assessed for demographics; significant differences between responders and non-respondersEffect estimate (Prevalence difference ) was adjusted for age, sex, race, branch of military and rankOverall risk of bias: low
Symptoms of alcohol abuse (NG/Res)
19.4 16.8Prevalence difference (R) 2.3 -0.8, 5.4
Prevalence difference (NG/Res)
2.6 -0.1, 5.3
ORa (all subjects)(males and females)
1.33 1.11, 1.58
Goss Gilroy 1998 (7)
Cross-sectional postal survey; 1997
All Canadian GWV (sea, land, air service; n = 3113) and sample of Canadian forces personnel eligible for active duty but non-deployed (n = 3439), matched on gender, age, regular/ reserve status
Defined by alcohol use in past month and ≥1 positive response in CAGE and of 5 questions on consequences of alcohol drinking in past 6 months
Symptoms suggestive of alcohol abuse
9.6(SD= 0.5)b
8.5(SD= 0.6)b
ORa
(males and females)
1.14 0.97, 1.35Participation rates: 73% GWV (4262 eligible); 60% non-GWV (5699 eligible)Non-response bias not assessed. A comparison between GWV and non-deployed personnel indicated no significant differences on socio-demographic and possible confounding factors.Adjusted prevalence ORs for symptoms suggestive of alcohol abuse were not reported.Overall risk of bias: low
GWV 93.9% males and 6.1% females (n=3113), non-GWV 94.2% males and 5.8% females (n=3439)
Steele 2000 (29)
Cross-sectional
Stratified random sample
Self-reported physician
Alcohol or drug 3.0 2.0
adj OR (males and females) 1.47 0.65, 3.31
Participation rate: 65% overall (of 3,138 original
86
Author Year (Reference)
Study design; study period
Sample Alcohol / substance use case defn measure
Outcome GWV prevalence
(%)
Comp group prevalence
(%)
Effect measure 95% Confidence Interval
Participation rates, Assessment of overall risk of bias
telephone interview; 1998
of US GWV residing in Kansas (n = 1548) and non-deployed comparison group (n = 482)
diagnosed alcohol or drug dependence new onset 1990-98 in males and females
dependence sample) and 93% GWV vs 88% non-GWV eligible in-state and contactable. 15% GWV’s military records indicated they had not served in the GW but reported that they had (excluded from analyses).Non-response bias assessed for demographics; significant differences between responders and non-responders on age, service branch, rank, sex.Effect estimate adjusted for age, sex, income and education levelOverall risk of bias: high
GWV 86% males and 14 % females (n=1548), non-GWV 87% males and 13% females (n=482)
McCauley 2002 (64)
Cross-sectional telephone interview; 1998 to 1999
Random sample of 3 groups of US Army or NG veterans living in 5 states and 1.) Serving within 50k of Khamisiyah Iraqi munitions site (KHAM GWV n=653), 2) Non-KHAM deployed (GWV other deployed n=610) and 3) Non-deployed (n=516)
Self-reported physician diagnosed alcohol or substance abuse diagnosed since the Gulf War
Alcohol or substance abuse
4.1(KHAM-deployed)3.6(non-KHAM deployed)
2.1 adj OR (KHAMnon-KHAM deployed combined) (males and females)
1.7 0.9, 3.4Participation rates: 78% KHAM GWV (838 eligible); 70% non-GWV (741 eligible)Non-response bias was assessed in the study for demographics, health outcomes; significant differences between responders and non-respondersEffect estimate adj for age, gender, race and region of residenceOverall risk of bias: high
Non-deployed group 89.3% males and 10.7% females (n=516), deployed group (KHAM and non-KHAM combined ) 95.1% males and 4.9% females (n=1263)
Ikin 2004 (1) Cross-sectional in-person structured diagnostic
All Australian male GWV (n=1381) and random sample of non-deployed
CIDI using DSM-IV criteria to assess 12-month
Alcohol dependence / abuse
19.8 12.6adj OR (males) 1.5 1.2, 2.0 Participation rates: 81%
of 1808 eligible GWV; 57% non-deployed (2796 eligible). Psychological interview: 78% GWV;
Drug 3.7 1.8 adj OR (males) 1.9 1.1, 3.2
87
Author Year (Reference)
Study design; study period
Sample Alcohol / substance use case defn measure
Outcome GWV prevalence
(%)
Comp group prevalence
(%)
Effect measure 95% Confidence Interval
Participation rates, Assessment of overall risk of bias
interview administered by trained psychologists; 2000-02
active duty personnel (n = 1377) matched by age, gender and service (Navy, Army, Air Force)
prevalence substance use disorder first present post-Gulf War in males
dependence / abuse
51% non-deployed. Participation bias assessed for demographics and health outcomes: significant differences between responders and non-responders on demographics but not on health outcomes.OR were adj for service, rank, age, education and marital status.Overall risk of bias: low
Any substance use disorder
20.8 13.1 adj OR (males) 1.5 1.2, 2.0
GWV 98% males and 2% females (n=1871), non-GWV 97.5% males and 2.5% females (n=2924)
Fiedler 2006 (66)
Cross-sectional telephone structured diagnostic interview administered by trained interviewers; 2000-01
Random sample of all US GWV (n = 967) and non-deployed era personnel (n = 784)
CIDI-Short Form using DSM-IV criteria. 12-month prevalence in males and females
Alcohol dependence 4.6 3.1
ORa (males and females)
1.51 0.91, 2.49 Participation rates: 59% GWV (1651 eligible); 51% Era non-deployed (1552 eligible). Non-response bias assessed for demographics; significant differences, with increased Whites and NCOs responding.OR for alcohol use or drug dependence were not adjusted. OR for combined alcohol/drug dependence in GWV or Era non-deployed deployed to a conflict in addition to or other than the GW; OR 1.91 (1.04, 3.54) were adj for rank, branch, sex, marital, and education.Overall risk of bias: low
Drug dependence 1.2 0.1
ORa (males and females)
9.84 1.28, 75.83
Any dependence
5.1 3.2 ORa,c (males and females)
1.62 0.99, 2.64
GWV 92% males and 8% females (n=967), Era veterans 88% males and 12% females (n=784)
Murphy 2006 (84)
Cross-sectional postal survey, not framed in
Stratified random sample of UK Armed forces services (Army, Navy,
AUDIT questionnaire. Q1 and Q2 modified to include a
Alcohol use disorders 7.4 6.6
ORa (males and females)
0.95 0.46, 1.94 Participation rate: overall approximately 57% (out of 2873 in original sampling frame were linked on GWV status)
88
Author Year (Reference)
Study design; study period
Sample Alcohol / substance use case defn measure
Outcome GWV prevalence
(%)
Comp group prevalence
(%)
Effect measure 95% Confidence Interval
Participation rates, Assessment of overall risk of bias
context of Gulf War deployment rather of a study to develop and evaluate screening questionnaires in the military2001
RAF) by service and size of unit, 45 individuals randomly selected from each unit. GWV status determined by record linkage. GWV (n = 308), non-GWV (n=1339)
higher category of units consumed
Non-response bias not assessed. Comparison between GWV and non-GW personnel indicated significant differences on service with greater proportion of Army, fewer Navy and RAF and slightly greater proportion males in the GWV groupEffect estimate adj for age, gender, rank, and serviceOverall risk of bias: low
GWV 99.4% males and 0.6% females (n=1339), NGW 95.2% males and 4.8% females (n=308)
Toomey 2007 (67)
Cross-sectional in-person structured diagnostic interview, by trained interviewers; 1998-2001
Stratified random subsample of US GWV (Navy, Army, Air Force, Marines (n = 1061) and non-deployed personnel (n = 1128) from a previous 1995 study
CIDI using DSM-IV criteria to assess substance dependence of Gulf War-era onset (1991-93)
Alcohol dependence 4.3 3.0
adj OR (males and females)
1.05 0.51, 2.16 Participation rates: 53% GWV (1996 eligible); 39% non-deployed (2883 eligible)Non-response bias was assessed. Significant differences on demographics but not health outcomesOR adj for age, gender, ethnicity, education, duty status (active vs reserve/guard), service, rank (illicit substance not adj for age or rank)Overall risk of bias: low
Illicit substance dependence
0.9 0.6adj OR (males and females)
1.20 0.27, 5.39
Any substance dependence 7.9 4.8
adj OR (males and females)
1.25 0.73, 2.16
Deployed 78% males and 22% females (n=1061), non-deployed 78% males and 22% females (n=1128)
Kang 2009 (31)
Cross-sectional postal and telephone
Follow-up stratified random sample from previous study
PHQ-9 criteria to assess probable alcohol abuse
Probable alcohol abuse 16.4 12.0
adj RRd (males and females) 1.24 1.11, 1.37
Participation rates: 40% GWV (15,508 eligible); 27% non-GWV (14,494 eligible).
89
Author Year (Reference)
Study design; study period
Sample Alcohol / substance use case defn measure
Outcome GWV prevalence
(%)
Comp group prevalence
(%)
Effect measure 95% Confidence Interval
Participation rates, Assessment of overall risk of bias
survey; 2004 of US GWV (Navy, Army, Air Force, Marines; n = 6111) and non-deployed Gulf Era personnel frequency matched on gender, service branch and status (n = 3859)
in past 6 months
Non-response bias assessed for demographics, health outcomes; significant differences between responders and non-responders on demographics but not on health outcomesEffect estimate adj for age, gender, race, BMI, current smoking, rank, service branch, unit component (active duty, national guard/reserve)Overall risk of bias: low
GWV 79.9% males and 20.1% females (n=6111), Era veterans 78.2% males and 21.8% females (n=3859)
AUDIT, World Health Organization Alcohol use disorder identification test (99); BMI, Body Mass Index; CAGE,CAGE Questionnaire (100); CIDI, World Health Organization Composite
International Diagnostic Interview (38); CIDI –SF, World Health Organization Composite International Diagnostic Interview Short Form (101); GW, Gulf War; GWV, Gulf War veterans;
NCO, Non-commissioned officer; NG, National Guard ; NG/Res, National Guard/Reserve; Non-deployed /non-GWV, a military comparison group who were not deployed to the Gulf War
during the period of operations. OR, Odds Ratio; PRIME-MD PHQ, Primary Care Evaluation of Mental Disorders Patient Health Questionnaire (102); R, Regular military; RR, Relative Risk;
SD, Standard Deviationa Unadjusted OR - calculated in Meta-XL (58), using reported prevalence and sample sizes for GWV and non-deployed personnel.b With no other theatre experience, defined as areas where Canadian Forces personnel served during past 12 years as part of a UN deployment.c Combined adjusted alcohol/drug dependence OR were only given for either GWV or non-GWV deployed to a conflict in addition to or other than the GW.d The unadjusted effect estimate was converted to an unadjusted OR in the meta-analysis.
90
Table 9 Characteristics of eligible studies comparing prevalence of alcohol/substance use disorders in Afghanistan/Iraq War veterans and non-deployed military personnel
First Author Year (Ref)
Study design study period
Sample Alcohol / substance use case definition / measure
Outcome Iraq/ Afghanistan veteran prevalence (%)
Comparison group prevalence (%)
Effect measure 95% Confidence Interval
Comments and assessment of overall risk of bias
Bray 2006 (69)
Cross sectional self-report anonymous questionnaire survey; 2006 (9th data point in series since 1980)
Sample of all eligible active duty US military personnel (excluded recruits, academy students, AWOL, or had permanent change of station). n=16,146 (3639 Army, 4627 Navy, 3356 Marine, 4524 Air Force)
AUDIT score >=20 assessed possible alcohol dependence. Questions assessed heavy alcohol use (≥5 drinks same occasion on ≥once/week past 30 days); any illicit drug use (past year)
Served in OEF/OIF vs non-deployed any operation- last 3 years: Alcohol dependence
3.52 2.29 Unadj OR 1.54a
1.1, 2.14 Overall response rate: 51.8%.Participants were selected to represent men and women in all pay grades of the active force worldwide. Data were weighted to represent all active duty personnel in the analyses.Overall risk of bias: low
Heavy alcohol use
19.4 16.69 Unadj 1.20 1.00, 1.45
Illicit drug use- Army
14.53 13.57 Unadj 1.09a 0.89, 1.33
Illicit drug use- Marine
11.58 8.87 Unadj 1.33a 0.93, 1.9
Illicit drug use- Navy
11.61 8.12 Unadj 1.52a 0.57, 3.46
Illicit drug use-Air Force
6.63 6.81 Unadj 0.97a 0.56, 1.62
Bray 2009 (70)
Cross sectional self-report anonymous questionnaire survey; 2008 (10th data point in series since 1980, included active duty Coast Guard for first time)
Sample of all eligible active duty US military personnel (excluded recruits, academy students, AWOL or incarcerated personnel). n=28,546; (5927 Army, 6637 Navy, 5117 Marine, 7009 Air Force, 3856
AUDIT score >=20 assessed possible alcohol dependence. Questions assessed heavy alcohol use (≥5 drinks same occasion on ≥once/week past 30 days); any illicit drug use (past year)
Combat deployed since Sept 11 2001 and served in OEF/OIF vs Not combat deployed since Sept 11 2001- Possible alcohol dependence past year
4.75 2.73 Unadj 1.78a 1.28, 2.45 Overall response rate: 71.6%.Participants were selected to represent men and women in all pay grades of the active force worldwide. Data were weighted to represent all active duty personnel in the analyses.Overall risk of bias: low
Heavy alcohol use
24.71 17.12 Unadj OR 1.58 1.26, 1.98
Illicit drug use- 28.14 24.88 Unadj OR 1.18a 0.91, 1.54
91
First Author Year (Ref)
Study design study period
Sample Alcohol / substance use case definition / measure
Outcome Iraq/ Afghanistan veteran prevalence (%)
Comparison group prevalence (%)
Effect measure 95% Confidence Interval
Comments and assessment of overall risk of bias
Coast Guard). ArmyIllicit drug use- Marine
24.07 20.84 Unadj OR 1.20a 0.93, 1.54
Illicit drug use- Navy
19.67 16.93 Unadj OR 1.20a 0.95, 1.53
Illicit drug use-Air Force
15.5 13.62 Unadj OR 1.16a 0.93, 1.45
Hotopf 2006 (90)
Cohort study self-report questionnaire; 2004-06
Stratified random sample of UK armed forces personnel deployed to Iraq Jan to April 2003 (Op TELIC 1) (n=4722, 3936 regulars, 786 reservists) and UK armed forces personnel at the time but not deployed to TELIC 1 (Era) (n=5550, 4750 regular, 800 reservists).
AUDIT to assess alcohol consumption and harmful use.Case defined as males with a score >13, females >10
Original TELIC 1 cohort AUDIT case
25.6 21.1 Adj OR 1.10Unadj 1.28
0.991.17
1.221.41
Participation rates: 62.3% deployed, 56.3% non-deployed samples.Non-response bias assessed via intensive follow up study, late responder analysis, sensitivity analysis, healthy warrior effect considered – unlikely important difference missed.OR adjusted for age, sex, rank, educational and marital status, service branch, fitness to deploy (and reservist status in regular vs reservist comparisons and take account of sampling weights).Overall risk of bias: low
Regulars 27.0 22.4 Adj OR 1.10 0.98 1.22Reservists 17.9 13.7 Adj OR 0.80 0.44 1.44
Iraq War vs Era AUDIT casea
27.0 21.2 adj OR 1.09a
Unadj OR 1.370.971.24
1.221.52
Combat vs non combat regulars in Iraq War group only
33.2 24.6 Adj OR 1.19Unadj OR 1.52
1.011.32
1.411.76
TELIC 1 sample 8% females, Era 10% females. Males and females included in analyses.No interaction by regular/reservist status (p=0.7) for AUDIT casenessaDistribution of main outcomes after reassigning Era participants to Iraq War group if they had served in subsequent TELIC 1 deployment.
Jacobson 2008 (103)
Follow up questionnaire survey of US Millennium
US Millennium Cohort Study participants who completed both
Patient Health Questionnaire (PHQ) assessed new onset alcohol
Active Duty OEF/OIF deployed with combat
4.8 3.6 Adj OR 1.03a 0.85 1.26 Follow-up response rate: 71.4%.Previous evaluations of possible biases indicate
92
First Author Year (Ref)
Study design study period
Sample Alcohol / substance use case definition / measure
Outcome Iraq/ Afghanistan veteran prevalence (%)
Comparison group prevalence (%)
Effect measure 95% Confidence Interval
Comments and assessment of overall risk of bias
Cohort Study; participants at baseline 2001-03 (n=77,047) and follow up2004-06 (n=55,021)
time point surveys and first deployment to Iraq or Afghanistan completed between baseline and follow up. 48,481 in analysis (Active duty=26,613; Reserve/NG=21,868). Of these, Deployed with (n=5510) or without (n=5661) combat. Non-deployed n=37310)
related problems (≥1 item). CAGE assessed alcohol dependence. Heavy drinking (males >14, females >7 drinks/week in previous week. Binge drinking (males ≥5, females ≥4 drinks on ≥1 day previous week or ≥5 drinks on ≥1 day/ occasion past year.
exposure vs non-deployed- PHQ Alcohol problems
cohort is representative sample military personnel, by demographic, mental health, health and exposure reporting. OR adj for sex, birth year, race/ethnicity, service, deployment length, history of mental disorders, smoking, service component, history potential alcohol dependence.
Overall risk of bias: low
Heavy weekly drinking
6.0 4.8 Adj OR 1.12 0.94 1.33
Binge drinking 26.6 19.3 Adj OR 1.31 1.14 1.49Reserve/Guard OEF/OIF deployed with combat vs non-deployed- PHQ Alcohol problems
7.1 3.8 Adj OR 1.63a 1.33 2.01
Heavy weekly drinking
8.8 5.1 Adj 1.63 1.36 1.96
Binge drinking 25.6 17.1 Adj OR 1.46 1.24 1.71Follow up survey respondents: deployed with combat exposure 8.1% males and 4.5% females, deployed without combat exposure 8.0% males and 5.5% females, and non-deployed 46.0% males and 55.1% females
Kline 2010 (75)
Cross sectional anonymous self-administered pre-deployment survey; 2007-08
New Jersey (NJ) National Guard (NG) members (n=2543) undergoing pre-deployment for Iraq: Two groups; ≥1 prior OEF or OIF deployment since 2001 (n=625) and No prior OEF/OIF deployments (n=1910).
Alcohol use measures based on questions/ algorithms in DSM-IV- based National Household Survey of Drug Use and Health. One measure of illicit drug use.
Alcohol dependence last 12 months
9.0 6.6 adj OR 1.88a 1.31, 2.69 Participation rate: 95% overall (of 2665 in original study popn)Non response bias not assessed due to survey anonymity. NJ NG included greater proportion of Hispanics, fewer non-Hispanic whites than a national NG sample, though few differences in current data on race. Similar to national sample on heavy drinking and 12 month
Binge drinking last 6 months
7.0 4.7 adj OR 2.29 1.51, 3.48
Heavy drinking last 12 months
19.5 20.3 adj OR 1.25 0.97, 1.62
Illicit drug use last 12 monthsa
10.0 11.6 adj OR 0.91a 0.66, 1.26
Nearly 25% previously deployed to OEF/OIF. Deployed 85.5% males, 14.5% females; non-deployed 89.0% males and 11.0% females. Males and females included in the analyses
93
First Author Year (Ref)
Study design study period
Sample Alcohol / substance use case definition / measure
Outcome Iraq/ Afghanistan veteran prevalence (%)
Comparison group prevalence (%)
Effect measure 95% Confidence Interval
Comments and assessment of overall risk of bias
drug use. Current sample 50% of NJ NG.Effect estimate adj for demographics, income, previous service in other conflicts.Overall risk of bias: low
Fear 2010 (3)
Cohort study questionnaire; 2007-09. Multiple phase study -2004-06 sample was reassessed and 2 new samples assessed
Randomly sampled UK armed forces (Army, Navy, Air Force), regulars (n=8278) and reservists (n=1712: included follow up of 2003 Iraq cohort (n=4203) (90), and two randomly sampled groups: deployment to Afghanistan only (n=1123) and Deployed to Iraq and Afghanistan (n=1389). Not deployed to Iraq or Afghanistan (n=3255).
10-item AUDIT. Score ≥16 defined hazardous use harmful to health- termed alcohol misuse
Alcohol Misuse – Regulars vs Non deployed
Iraq or Afghanistan
15.7 10.9 adj ORa 1.22 1.02, 1.46 Participation rate: 56% of eligible sampleNon response bias assessed: Mental health at stage 1 not associated with participation Sample and response weights applied in analysesOR adj for age, sex, marital, education, rank, serving status and serviceOverall risk of bias: low
Iraq only 15.3 10.9 adj OR 1.21 0.99, 1.46
Afghanistan only 17.8 10.9 adj OR 1.20 0.92, 1.57
Both Iraq and Afghanistan
15.1 10.9 adj OR 1.30 1.01, 1.67
Regulars: Combat vs combat service support
22.5 14.2 Adj OR 1.15 0.93, 1.42
Combat support vs combat service support
10.8 14.2 Adj OR 0.68 0.50, 0.93
Iraq or Afghanistan 9.5 6.8 Adj ORa 1.38 0.89, 2.13
Deployed to Iraq 90.6% males, 9.4% females; Afghanistan only 94.5% males, 5.5% females; deployed to Iraq and Afghanistan 94.3% males, 5.7% females (n=1389); not deployed to Iraq or Afghanistan 86.8% males, 13.2% females (n=3255).
Shen 2012 (78)
Database analysis on several
678 382 unique active duty personnel
ICD-9 diagnosis of any substance use/ dependence
Any substance use/ dependence
Army 14.8 6.0 Adj OR 4.05 3.82, 4.30 Sample 25% and representative of US Armed Forces active duty
Marines 9.3 5.0 Adj OR 4.36 3.82, 4.97Navy 8.6 8.0 Adj OR 1.77 1.45, 2.16
94
First Author Year (Ref)
Study design study period
Sample Alcohol / substance use case definition / measure
Outcome Iraq/ Afghanistan veteran prevalence (%)
Comparison group prevalence (%)
Effect measure 95% Confidence Interval
Comments and assessment of overall risk of bias
combined sources from US Defense Manpower Data Center and TRICARE, to obtain active duty personnel, demographic service, mental and related health information
serving b/n 2001 and 2006, approx 49% Army, 14% Marine, 20% Navy, 17% Air Force. Four groups: Not deployed under OEF/OIF; Deployed- Iraq or Afghanistan; Other known locations under OEF/OIF, e.g. Kuwait, Qatar, Turkey; Classified/ unknown locations
disorder between 2001 and 2006; according to DSM-IV criteria. Also considered 3 deployment lengths (data not shown)
Deployed Iraq or Afghanistan vs Not deployed
population, Captured diagnoses from in and outpatient settings and civilian and military providers.Undertook sensitivity analyses. OR adjusted for demographic (gender, race/ethnicity, marital age) and service characteristicsOverall risk of bias: low
Air Force 7.1 5.9 Adj OR 1.76 1.56, 1.99
OEF-OIF deployed females - Army 11% Marines 3%, Navy 13%, Air Force 15% and similar to Not deployed.
Vanderploeg 2012
Cross sectional anonymous
Current active members of the Florida National
AUDIT identifying excessive drinking behaviours. Cutoff
Excessive drinking
38.4 29.8 adj OR (males and
0.75a 0.57, 1.00 Participation rate: 41.3% (9700 eligible)
95
First Author Year (Ref)
Study design study period
Sample Alcohol / substance use case definition / measure
Outcome Iraq/ Afghanistan veteran prevalence (%)
Comparison group prevalence (%)
Effect measure 95% Confidence Interval
Comments and assessment of overall risk of bias
(79) online survey; 2009-10
Guard who were deployed to OEF/OIF (n=1443) and not deployed (n=1655)
ref to Hoge (74) which measured alcohol misuse using a 2-item screening instrument (104)
females) Effect estimate adj for other covariates, demographic, pre-deployment factors
Non response bias not assessed due to survey anonymity, No information on Florida NG cohort available Overall risk of bias: high
Deployed 87.4% males, 12.6% females, non-deployed 79.2% males, 20.8% females
Trautmann 2014 (98)
Cross sectional clinical- epidemiological study including face-to-face diagnostic interviews; 2010-2011
German soldiers examined about 12 months after deployment to Afghanistan in 2009/2010 (n=1483) and randomly selected never deployed soldiers stratified by age, sex and unit grid (n=889)
Munich-CIDI using DSM-IV criteria to assess AUD, and to derive binge drinking (≥7 drinks on 1 occasion) and heavy drinking (≥24g for men or ≥16g for women of ethanol /day)
Any AUD last 12 months
3.6 2.2 adj OR (males and females)
1.9a
1.0
1.1
0.6a
0.99, 3.5
(0.7-1.3)
(0.8-1.6)
(0.2-2.0)
Participation rate: 92.8% (1599 eligible deployed Afghanistan) vs 95.4% (932 eligible never deployed)Effect estimate adj for demographics, economic situation, service length, unit, rank.Eligible and non-eligible soldiers similar. Never deployed and deployed comparableOverall risk of bias: low
Binge drinking 36.2 38.1Heavy drinking 13.9 13.9SUD 1.0 1.7Deployed and never deployed both 95.0% males
AUDIT, World Health Organization Alcohol use disorder identification test (99); CAGE,CAGE Questionnaire (100); CIDI, World Health Organization Composite International Diagnostic Interview (38); NG, National Guard ; Non-deployed /non-Afghanistan/Iraq, a military comparison group who were not deployed to Afghanistan/Iraq War during the period of operations. OR, Odds Ratio
a OR data used in the meta-analyses
96
4.6.2 Key findings
Nine studies in Gulf War veterans and nine studies in Afghanistan/Iraq War veterans met the
inclusion criteria. Pooled analysis based on random-effects model indicated that Gulf War
veterans and Afghanistan/Iraq War veterans were at higher alcohol use disorder risk than military
personnel not deployed to these conflicts respectively, with the summary ORs being very similar
for the two groups. The summary estimate of the likelihood of alcohol use disorders was slightly
higher, but not statistically significant, in reservists compared with regular personnel deployed to
Afghanistan/Iraq War. Veterans of the Afghanistan/Iraq War were at increased risk of substance
use disorders, but it was difficult to draw definitive conclusions regarding substance use disorders
in Gulf War veterans since only three studies were available reporting this outcome and the
elevated meta-analytic estimate had very wide confidence intervals. Our meta-analysis also
suggested that Gulf War veterans were at a significantly increased risk for any substance use
disorder compared with military personnel who were not deployed to the Gulf War. The results
from alcohol and any substance use disorder outcomes were robust to the impact of risk of bias,
publication bias and our sensitivity analyses.
97
4.7 Generalised anxiety disorder in Gulf War and Afghanistan/Iraq War veterans compared to non-deployed military personnel
4.7.1 Results
Of the 25 primary Gulf War veteran health studies identified, five studies were eligible to be
included in the systematic review and meta-analysis in relation to GAD. Of the 24 primary
Afghanistan/Iraq War veteran health studies identified, three studies reported on GAD and were
included in the meta-analysis. The characteristics of these studies are reported in Table 10 and
Table 11.
Across the five included Gulf War studies, sample sizes ranged from 204 to 1896 for Gulf War
veterans and 48 to 1799 for the non-deployed personnel; and across the three included
Afghanistan/Iraq War veteran studies sample sizes ranged from 1443 to 5325 for Afghanistan/Iraq
War veterans and from 1655 to 4886 for the non-deployed personnel. The percentage of female
veterans in these studies ranged from 1% to 22%.
All except one of the studies was based on US veteran cohorts, the other study population was
Australian Gulf War veterans. These studies included personnel from all services (Army, Navy,
Marines, Air Force, National Guard). The military comparison groups in all the studies were
described and defined as non-deployed, i.e. not deployed to the Gulf War or Afghanistan/Iraq War
during the period of operations, rather than other conflict personnel. Some studies attempted to
oversample certain groups such as women and reservists to determine whether these groups had
different rates of GAD (17). Most studies did not endeavour to sample specific categories of
military personnel in order to mirror deployment proportions, though samples were generally
representative of their populations.
Of the Gulf War studies, four used structured diagnostic interviews (CIDI or SCID) and one used a
screening tool (PRIME-MD). For the outcome of GAD, two provided data for adjusted OR in the
published article. The other three studies provided usable data in the published article for
calculation of an unadjusted OR. Of the Afghanistan/Iraq War veteran studies, all three studies
used screening tools (PHQ or 7Q-GAD scale). Two provided adjusted ORs, and the third provided
an adjusted prevalence. The authors of this study had been contacted and had provided on
request more precise estimates of standard errors of prevalence that enabled calculation of ORs.
98
Gulf War veteran studies reporting GAD
The overall OR using the random effects model for the five included studies was 3.04 (95% CI:
1.95, 4.75) which indicated a significantly higher risk of GAD in Gulf War veterans compared to
non-deployed military personnel. Overall heterogeneity for all studies represented by I2 was low to
moderate at 35% (Figure 26). Sensitivity analyses indicated that the overall OR did not vary much
after excluding any individual study of the five included studies. Black et al. (17) reported ORs
separately for reservists and regular military personnel. These two subpopulations were also
entered separately into the meta-analyses. Excluding the reservists’ subgroup caused the OR to
decrease the most to 2.40 (95% CI: 1.65, 3.49).
Figure 26 Random-effects meta-analysis illustrating log-transformed odds ratios of generalised anxiety disorder in Gulf War veterans and non-deployed military personnel
Stratification by case definition indicated that the diagnostic interview subgroup (OR = 2.48, 95%:
1.49, 4.10) and screening tool subgroup (OR = 3.40, 95%: 1.59, 7.27), with only one study in it but
two subpopulations, had broadly similar ORs and confidence intervals. Heterogeneity was 0% in
the diagnostic interview subgroup but high and statistically significant in the screening tool
99
subgroup (I2=73%, p=0.05); the screening tool subgroup contributed more weight in calculating the
overall OR (58.68%) (Figure 27).
Figure 27 Random-effects meta-analysis of generalised anxiety disorder in Gulf War veterans and non-deployed military personnel, subgrouped by case definition
The studies were further stratified based on risk of bias, and the OR in both subgroups were
comparatively similar (low risk of bias subgroup OR = 3.34, 95% CI: 1.86, 5.99; high risk of bias
subgroup OR = 2.89, 95% CI: 0.77, 10.94). Heterogeneity was 47% in the low risk subgroup and
33% in the high risk subgroup, but both figures were not at the level of significance. The low risk
subgroup contributed substantially more weight to the overall OR (Figure 28).
Three of the five studies received a high overall risk of bias. Common factors contributing to this
assessment were poorer sample designs (i.e. convenience samples, non-random sample), lack of
adjustment for possible confounding factors, high non-response rates and lack of calculation of
ORs. All of the studies that did not report adjusted ORs were assessed as having an overall high
risk of bias.
100
Figure 28 Random-effects meta-analysis of generalised anxiety disorder in Gulf War veterans and non-deployed military personnel, subgrouped by risk of bias
Afghanistan/Iraq War veteran studies reporting GAD
The Hoge et al. (74) study reported ORs for three distinct subpopulations based on theatre of
operation (OEF, Afghanistan or OIF, Iraq) and service branch (Army or Marines). Thus these
subpopulations were combined using Stata (version 11) and entered separately into the meta-
analysis, in order to avoid triple counting the non-deployed group. It also reported ORs using
either a strict or broad definition of GAD based on the Patient Health Questionnaire (PHQ)
screening tool, where the strict case definition required reporting of functional impairment to make
a positive diagnosis of GAD, and the broad case definition focussed only on reported symptoms for
a positive diagnosis. The strict PHQ definition of GAD was used in order to limit heterogeneity in
the meta-analysis, as it was more similar to the case definitions used by Vanderploeg et al. (79)
and Bray et al. (91).
The overall OR using the random effects model for the three included studies was 1.20 (95% CI:
1.00, 1.44) which indicated a statistically significantly higher risk of GAD in Afghanistan/Iraq War
101
veterans compared to non-deployed military personnel. Heterogeneity was detected at 0% (Figure
29).
Figure 29 Random-effects meta-analysis illustrating log-transformed odds ratios of generalised anxiety disorder in Afghanistan/Iraq War veterans and non-deployed military personnel
Subgroup analyses by risk of bias, case definition or any other factor could not be done due to the
low number of studies. Both Hoge et al. (74) and Bray et al. (91) were assessed as low risk of bias
whereas Vanderploeg et al. (79) was assessed as high risk of bias due to a low study response
rate and inability to assess non-response bias. However, sensitivity analyses indicated that the
overall OR did change and become non-significant after excluding either the Bray et al. (91) study
or Hoge et al. (74) study. The Bray (2006) study was completed as part of a US Department of
Defense survey solely on reservists (from all service branches) and National Guard personnel. For
this reason the study was removed to ascertain the effect on the meta-analysis: the OR decreased
to 1.14 (0.94, 1.40) and became non-significant (Figure 30).
102
Odds Ratio
Figure 30 Random-effects meta-analysis of GAD in Afghanistan/Iraq War veterans compared to non-deployed military personnel, sensitivity analysis excluding Bray (2006)
Comparison of summary odds ratio of generalised anxiety disorder in Gulf War veterans versus Afghanistan/Iraq War veterans
Furthermore, the summary OR of GAD in Gulf War veteran studies (OR 3.04, (95% CI 1.95, 4.75)
was statistically significantly higher than the summary OR (OR 1.20, 95% CI 1.00, 1.44) for the
Afghanistan/Iraq War veteran studies, as their respective confidence intervals did not overlap.
103
Table 10 Characteristics of eligible studies comparing prevalence of generalised anxiety disorder (GAD) in Gulf War veterans and non-deployed military personnel
First author
Study design, study period
Sample GAD case definition and measure
Outcome GWV prevalence
(%)
Comp group
prevalence (%)
Odds Ratio (95% CI)
Participation rates; Assessment of overall risk of bias
Black 2004 (17)
Cross-sectional, structured telephone interview; 1995 to 1996
Random sample GWV selected from Iowa US; regular military (Reg) (n = 985), GW National Guard (NG)/Reserve (Res) (n = 911), non-GW regular military (n = 968), and non-GW NG/Res (n = 831)
PRIME-MD, based on the DSM-IV; GAD required 3 or more symptoms of anxiety for past year and anxiety/worry more days than not during past 12 months
12 month GAD (Reg)*
3.9 1.9 2.3 (1.3-4.0) Study participation rates: 91% (76% of eligible)Non-response bias assessed in study: noOverall risk of bias: low
12 month GAD (NG/Res)*
4.5 1.0 5.0 (2.9-8.8)
12 month GAD (total)
4.0 1.8 2.5 (1.5-4.1)
Prevalences weighted to account for complex study design; ORs adjusted for age, sex, race, rank, branch of service, military status, and pre-GW mental health conditionsNumber of females in sample not reported
Wolfe 1999 (6)
Cross-sectional; in-person structured clinical interview by trained clinicians; 1994 to 1996
Stratified, random sample two cohorts US GWV from Fort Devens, New England (FD, n = 148) and New Orleans, (NO, n = 56), comparison group of air ambulance unit deployed to Germany during GW (G, n = 48)
SCID non-patient edition to assess current (within 1 month) GAD disorder
1 month GAD (FD)
0.8 0.0 Study participation rates: 62% (FD; 353 eligible); 38% (NO; 194 eligible); 85% (G; eligible numbers not reported)Psychological interview participation rates: 42% (FD), 30% (NO), 51% (G)Non-response bias assessed in study: yes, for demographics, health outcomesSignificant differences between respondents and non-respondents: yes, on demographics and health outcomesOverall risk of bias: high
1 month GAD (NO)
0.0 0.0
1 month GAD (total)*
0.8 0.0 0.71 (0.03-17.82)a
Prevalences adjusted for stratification variables of gender and health symptomsFemales made up 12.0% of total sample
Toomey 2007 (67)
Cross-sectional, in-person computerised diagnostic interview; 1998 to 2001,
Stratified, random subsample of previous study by Kang (1995). US GWV (n = 1061) and non-deployed (n = 1128)
CIDI using DSM-IV criteria to assess 12-month GW-era onset (Jan 1991-July 1993) GAD; Beck Anxiety Inventory
GW-era onset GAD
0.9 0.0 22.54 (1.32-385.1)a
Psychological interview completion rates: 53% GWV (1996 eligible) 39% non-GWV (2883 eligible)Non-response bias assessed in study: yes, for demographics, health outcomesSignificant differences between respondents and non-respondents: yes,
Females made up 22% of the total sample
104
First author
Study design, study period
Sample GAD case definition and measure
Outcome GWV prevalence
(%)
Comp group
prevalence (%)
Odds Ratio (95% CI)
Participation rates; Assessment of overall risk of bias
10 years after the war
assessed past week anxiety symptoms-total score indicating minimal (0-7), mild (8-15), moderate (16-25), severe (26-63) anxiety
on demographics; not on health outcomesOverall risk of bias: high
Ikin 2004 (1)
Cross-sectional, in-person structured diagnostic interview by trained psychologists; 2000 to 2002, 10 years after the war
All Australian GWV (n = 1381) and random sample of Navy, Army and Air Force non-deployed active duty personnel (n = 1377) matched by age, gender and service type
CIDI using DSM-IV criteria to assess post-Gulf War GAD
First present post-GW GAD*
0.7 0.2 2.9 (0.7-16.4) Participation rate: 81% GVW (1808 eligible); 57% non-GWV (2796 eligible)Psychological interview completion rates: 78% GWV; 51% non-deployedNon-response bias assessed in study: yes, for demographics, health outcomesSignificant differences between respondents and non-respondents: yes, on demographics; not on health outcomesOverall risk of bias: low
Pre- and post-GW GAD in previous 12 months
0.4 0.1 2.6 (0.5-27.0)
OR were adjusted for service type, rank, age, education and marital status or where numbers were small for service type, rank and ageFemales made up 2.4% of the total sample. Analyses were restricted to males only
Fiedler 2006 (66)
Cross-sectional, telephone structured diagnostic interview by trained interviewers; 2000 to 2001
Random sample (n = 967) of all US GWV and non-deployed personnel (n = 784)
CIDI-Short Form (SF) using DSM-IV criteria to assess 12 month GAD
12 month GAD (total)*
6.0 2.7 2.32 (1.37-4.06)a Participation rates: 59% GVW (1651 eligible); 51% non-GWV (1552 eligible)Psychological interview completion rates: 55% GWV; 43% non-deployedNon-response bias assessed in study: yes, for demographicsSignificant differences between respondents and non-respondents: yesOverall risk of bias: high
12 month GAD (males)
5.8 2.2
12 month GAD (females)
8.0 6.5
Females made up 9.6% of total sample
Gulf War, Gulf War veterans and non-deployed comparison group/s abbreviated as GW, GWV and non-GWV for brevity; a = Unadjusted OR (Calculated through Stata 11 and RevMan 5.3 using reported prevalences and sample sizes for deployed and non-deployed personnel); * Odds ratios that were used in the meta-analyses; OR = Odds Ratio; NG= National Guard; Res= Reserve; non-deployed = a military comparison group who were not deployed to the Gulf War during the period of operations; CIDI = World Health Organization Composite International Diagnostic Interview (38); SCID = Structured Clinical Interview for DSM Disorders; PRIME-MD PHQ = Primary Care Evaluation of Mental Disorders Patient Health Questionnaire based on DSM-III-R criteria (105)
105
Table 11 Characteristics of eligible studies comparing prevalence of generalised anxiety disorder (GAD) in Afghanistan/Iraq War veterans and non-deployed military personnel
First author
Study design and study period
Sample GAD case definition and measure
Outcome GWV prevalence
(%)
Comp group prevalence (%)
Odds Ratio (95% CI)
Participation rates; Assessment of overall risk of bias
Vander-ploeg 2012 (79)
Cross-sectional, anonymous online survey; 2009 to 2010
Sample from all active members of the Florida NG (n = 3098), comprising deployed group (n = 1443) and non-deployed group (n = 1655)
7-question Generalized Anxiety Disorder scale; reported nervousness or worry more than half the time, 3 of 6 other anxiety symptoms more than half the time, and impairment in functioning at the “very difficult” level.
Current GAD 4.2 1.5 0.70 (0.23-2.15) Study participation rates: 41.3% overall (10,400 eligible)Non-response bias assessed in study: no due to survey anonymityOverall risk of bias: high
OR adjusted for all other covariates, and demographic (sex, race, education level, marital status) and pre-deployment (prior psychological trauma or probable TBI) factorsFemales made up 17.0% of the total sample
Hoge 2004 (74)
Cross-sectional, in-person questionnaire; 2003
Samples from Army combat infantry brigade pre deployment to Iraq (n = 2530); an infantry brigade of same division post deployment to Afghanistan (n = 1962); an Army infantry brigade of post Iraq deployment (n = 894); and Marine Corp units post Iraq deployment (n = 815)
Patient Health Questionnaire (PHQ), modified slightly; two case definitions- broad screening definition not including criteria for functional impairment or for severity, strict screening definition requiring self-report of substantial functional impairment or a large number of symptoms
Current GAD (broad)
Army, Afghan.
17.2 15.5 1.13 (0.96-1.33) Non-response bias assessed in study: noStudy participation rates: 98% for the four samples combinedQuestionnaire completion rate: 95% for the anxiety and depression measuresThe demographic characteristics of the sample closely mirrored that of the populationOverall risk of bias: low
Current GAD (strict)
7.4 6.4 1.17 (0.92-1.48)
Current GAD (broad)
Army, Iraq
17.5 15.5 1.16 (0.94-1.43)
Current GAD (strict)
7.9 6.4 1.25 (0.92-1.68)
Current GAD (broad)
Marine Corp, Iraq
15.7 15.5 1.02 (0.81-1.27)
Current GAD (strict)
6.6 6.4 1.03 (0.74-1.43)
Current GAD (strict)*
Deployments combined 1.16 (0.94-1.43)
OR adjusted for differences in demographic characteristics before and after deployment (age, sex, race, education level, rank, marital status)Females made up 1.0% of the total sample
106
First author
Study design and study period
Sample GAD case definition and measure
Outcome GWV prevalence
(%)
Comp group prevalence (%)
Odds Ratio (95% CI)
Participation rates; Assessment of overall risk of bias
Bray 2006 (91)
Cross-sectional, self-report questionnaire survey; 2006 (9th data point in series since 1980)
Sample of all US reserve and NG military personnel (excluded recruits, academy students, AWOL or had permanent change of station) comprising army, navy, marine and air force reserve and coast guard. Served in OEF/OIF (n = 5325) and those non-deployed (n = 4886)
PHQ; if bothered by feelings of anxiousness for several days in the past month (30 days) and had at least three other symptoms for more than half the days, they were scored as needing further anxiety evaluation
1 month GAD 12.76 (SE 1.36) 9.13 (SE 0.76) 1.41 (0.97-2.01)a Overall response rate: 55.3%Non-response bias assessed in study: yes, a non-response adjustment was made to help compensate for the potential bias of nonsurveyed personsOverall risk of bias: low
Prevalences adjusted corrected for differences in the demographic distributions between the two groups. The main effect of reserve component, gender, age group, enlister/officer indicator, married/other, education and race/ethnicity were used in this standardisation processFemales made up 21.2% of the total sample
a = Unadjusted OR (Calculated through Stata 11 and RevMan 5.3 using reported prevalences and sample sizes for deployed and non-deployed personnel); * Odds ratios that were used in the meta-analyses; OR = Odds Ratio; NG= National Guard; OEF= Operation Enduring Freedom (Afghanistan); OIF= Operation Iraqi Freedom; non-deployed = a military comparison group who were not deployed to the Afghanistan/Iraq War during the period of operations.
107
4.7.2 Key findings
The summary OR using the random effects model for the five included studies in the meta-analysis
of GAD indicated Gulf War veterans were at three times the odds of GAD compared with non-
deployed military personnel. Overall heterogeneity for all studies represented by I2 was moderate
at 35%, but not statistically significant. This finding did not vary much after excluding individual
studies one by one in sensitivity analyses. In subanalyses the OR for GAD in Gulf War was lower,
but still increased over two-fold, in the subgroup assessed by diagnostic interview, and in subgroup
analysis when the reservist subgroup in one of the studies was excluded. The number of studies
was small however and these findings should be interpreted with caution.
The summary OR using the random effects model for the three included studies in the meta-
analysis indicated that Afghanistan/Iraq War veterans were at twenty percent increased odds of
GAD compared with non-deployed military personnel. Heterogeneity was not at a detectable level.
Further subanalyses could not be undertaken due to the small number of studies. However, when
a study that was undertaken solely on reservists and National Guard was excluded, the OR
decreased and was not statistically significant.
The difference in pooled OR between Gulf War veteran studies and Afghanistan/Iraq War veteran
studies was statistically significant, as their respective confidence intervals did not overlap.
108
5 Multisymptom illness in Gulf War, Afghanistan and Iraq War veterans compared to non-deployed military personnel
5.1 Literature search results
Figure 31 shows that the search yielded 2,573 records, with 2,445 records remaining after removal
of duplicates. The titles and abstracts were screened to identify studies for full- text review by the
specified inclusion and exclusion criteria. After abstract review, 130 full-text articles were identified
for further review, and seven eligible articles were identified reporting multisymptom illness
according to the eligibility criteria. Figure 31 also identifies the reasons for excluding full text
articles that were assessed for eligibility for multisymptom illness.
109
2,573 Records Identified Through Database Search and Additional Searching
7 Eligible Studies Reporting Multisymptom Illness (one
study had two articles)7 Gulf War veterans
0 Afghanistan/Iraq veterans
122 Full-text Articles Excluded because:2 Subanalysis of MSI4 Deployment was not specified or non-Gulf/Afghanistan/Iraq77 Inappropriate comparison group or case definition or no measure of MSI 38 Study not original research or odds ratio not calculable 1 Study based on treatment seeking sample
130 Full-text Articles Assessed for Eligibility
2,315 Abstracts Excluded
2,445 Abstracts Assessed For Eligibility (excluding
duplicates)
Figure 31 Flowchart of the systematic review of multisymptom illness (MSI) Gulf War veterans, Afghanistan/Iraq War veterans, 1990-2014
The systematic review identified seven studies of veterans from the 1990-1991 Gulf War, five from
the US, one from the UK and another from Australia. No studies of Iraq or Afghanistan veterans
were identified. Hence the rest of this section is based on Gulf War veterans alone.
The key features of the selected studies are summarised in Table 13 and the studies are arranged
according to the year in which data were collected. All studies were cross-sectional and the
methods of data collection used were similar across studies except that some studies used postal
questionnaires while other studies complemented this through telephone interviews. The study
populations in most studies included a range of military service branches, an exception was the
study by Fukuda et al. (27) which was conducted among US Air Force personnel only, while the
study by Kelsall et al. (1) invited all Australian Gulf War veterans to participate but the Gulf War
cohort was predominantly Navy personnel. The definition of multisymptom illness used in most
studies included was very similar to the three symptom groupings developed by Fukuda et al. (27)
(also termed as the CDC definition). Only the studies by Kelsall et al. (2006) and Kang et al.
(2009) (31) used modifications of the CDC definition. Kelsall et al. (2009) (1) multisymptom illness
definition included symptoms across multiple body systems and was based on the CDC definition,
but the symptom groupings were derived from factor analysis empirically derived in Australian Gulf
War veterans and included three symptom groupings rather than the two symptom groupings
identified by Fukuda et al. (27) Kang et al. (31) multisymptom illness definition used a multi-body
system-like definition although it was less methodically described.
Three studies were assessed as having overall high risk of bias and these were the studies by
Fukuda et al. (1998), Proctor et al. (2001) and Kang et al. (2009). The rationale for this
assessment included that in the study of Fukuda et al. (27) only serving members were included,
the response rate was moderate and unadjusted ORs only were available. The Devens cohort,
reported by Proctor et al. (2001) (106) was not considered representative of the target population
and had a moderate response rate with significant differences between participants and non-
participants (107). As mentioned above, the multisymptom illness definition as reported by Kang et
al. (2009) was less methodically described than others, but sufficiently to include, and the response
rate was relatively low.
110
5.2 Multisymptom illness in Gulf War veterans compared to non-deployed military personnel
5.2.1 Results
Meta-analysis of the seven eligible studies showed that the odds of multisymptom illness were
more than two and a half times greater in Gulf War veterans compared with the non-deployed
comparison groups (OR = 2.74, 95% CI 2.15, 3.51) (Figure 32). Figure 32 also presents the study-
specific effect sizes. Between–study heterogeneity was high (I2=92%).
OR54321
Study
Kelsall 2009
Blanchard 2006
Proctor 2001
Unwin 1999 & 2002
Overall
Q=70.25, p=0.00, I2=91%
Kang 2009
Steele 2000
Fukuda 1998
OR (95% CI) % Weight
1.80 ( 1.48, 2.19) 15.69
2.16 ( 1.61, 2.90) 13.95
2.40 ( 1.10, 5.30) 6.26
2.41 ( 2.10, 2.76) 16.54
2.74 ( 2.15, 3.51) 100.00
3.05 ( 2.77, 3.36) 16.98
3.26 ( 2.48, 4.28) 14.35
4.69 ( 4.00, 5.51) 16.23
Odds ratio
Figure 32 Random-effects meta-analysis of multisymptom illness in Gulf War veterans and non-deployed military personnel
Subgroup analyses were also conducted to further explore factors that could explain heterogeneity.
Analysis by risk of bias showed that the estimated summary OR among studies with low risk of
bias was OR 2.33 (95% 1.87, 2.91; I2=77%) and the estimate summary OR from studies with high
risk of bias was greater with an OR 3.51 (95% CI 2.41-5.13; I2=91%). The summary OR of the
high risk of bias studies was estimated 52% higher than estimated by the studies with low risk of
bias, but this was not statistically significant (OR = 1.52, 95% CI 0.89, 2.61).
111
Further subgroup analysis was conducted among Fukuda et al. and the four studies that based
their definition of multisymptom illness on the three groupings of symptoms used in the CDC
definition (excluding Kelsall et al. (2009) and Kang et al. (2009)). Subgroup analysis of
multisymptom illness in these five studies found that the odds of multisymptom illness was almost
three times greater among Gulf War veterans than the comparison groups (OR = 2.93, 95% CI
2.06, 4.18; I2=91%)
Sensitivity analyses, presented in Table 12, did not reveal any marked study specific effects with
the exception of the study reported by Fukuda et.al. (1998).
Table 12 Sensitivity analyses excluding each study one by one for studies of multisymptom illness in Gulf War veterans compared to non-deployed military personnel
Excluded study Overall
Pooled ES 95% CI I 2
Fukuda 1998 2.48 2.03 – 3.03 83%
Proctor 2001 2.77 2.14 – 3.58 93%
Steele 2000 2.66 2.02 – 3.51 93%
Unwin 1999 & 2002 2.81 2.08 – 3.78 92%
Blanchard 2006 2.85 2.18 – 3.73 92%
Kelsall 2009 2.98 2.35 – 3.78 89%
Kang 2009 2.68 1.91 – 3.75 93%
ES= effect size
Presented in Figure 33 is the funnel plot. A visual inspection of the funnel plot shows that six of the
seven studies were symmetrically distributed across the top of the funnel (with one smaller study
within the funnel at the base). The horizontal scatter around the summary fixed estimate is
suggestive of study heterogeneity. (Note: FE MA, Fixed effects meta-analysis)
112
Figure 33 Funnel plot for the random-effects meta-analysis of multisymptom illness Gulf War veterans and non-deployed military personnel
113
Table 13 Characteristics of eligible studies comparing prevalence of multisymptom illness in Gulf War veterans and non-deployed military personnel
First Author
Study design; study period
Sample Multisymptom illness (MSI)case definition*
GWV prevalence
(%)
Comp group prevalence
(%)
ES(Adj. OR)
95% CI Comments and Assessment of overall risk of bias
Fukuda, 1998(27)
Cross sectional study with self-administered questionnaires, 1995
US Air Force personnel stationed at 4 Air Force bases.N=1,163 GWVN=2,560 non-GWV
CDC definition: Presence, for 6 months or more, of at least one symptom from two or more symptom groupings namely (i)fatigue; (ii) mood/cognition (i.e. feeling depressed, difficulty remembering or concentrating, feeling moody, feeling anxious, trouble finding right words or difficulty sleeping); and (iii) musculoskeletal (i.e. joint pain/stiffness or muscle pain). The symptom groupings mood/cognition and musculoskeletal were derived through exploratory principal components analysis in this study group
45 15 Mild-Moderate: 4.08Severe: 16.18
No adjusted combined estimate
3.39-4.93
8.99-29.14
Participation rates: Participation in the study was 61% (3723/6151) and it is not presented separately for GWV and non-GWV)Non-response bias was assessed and concluded that demographic characteristics of participants were similar to those of their respective units.Factors adjusted for in estimation of OR were not listedOverall risk of bias: High
Proctor, 2001(106)
Cross sectional study with self-administered questionnaires, 1994-1996.
Gulf War and German (deployed during Gulf War) veterans from Ft Devens (Massachusetts, US).N=180 GWVN=46 non-GWV
CDC-derived definition: Presence, for 6 months or more, of at least one symptom from two or more symptom groupings namely (i) fatigue (i.e. fatigue or easily tired); (ii) mood/cognition (i.e. frequent periods of feeling depressed, forgetfulness, difficulty concentrating, crying easily, excessive anger or irritability, frequent periods of anxiety and nervousness, inability to fall asleep, restless or unsatisfying sleep or awake earlier than desired); and (iii) musculoskeletal (i.e. neck aches/ stiffness or joint pains). The symptoms also needed to have begun during or after the Gulf War
65.3 32.6 2.4 1.1-5.3 Participation rates: 62% in GWV & 51% in non-GWV.Non-response bias was assessed only among GWV and participants differed from non-participants with respect to sex, race/ethnicity, age, education and symptomatology as reported in the Phse-1 study. Other factors were similar (i.e. employment status, marital status, alcohol/drug use, or service status).(107)ORs adjusted for participation bias, age and psychiatric casenessOverall risk of bias: High
Steele, 2000 (29)
Cross-sectional
US GWV and non-GWV residing in
CDC-derived definition: Presence of at least one symptom from two or more
47.2 19.8 3.26 2.48-4.28 Participation rates: 93% GWV & 88% in non-GWV.
114
First Author
Study design; study period
Sample Multisymptom illness (MSI)case definition*
GWV prevalence
(%)
Comp group prevalence
(%)
ES(Adj. OR)
95% CI Comments and Assessment of overall risk of bias
telephone interview; 1998
Kansas and non-deployed comparison groupN = 1,548 GWVN = 482 non-GWV
symptom groupings namely (i) fatigue; (ii) mood/cognition (i.e. feeling down/depressed, memory problems, difficulty concentrating, trouble finding words, problems falling or staying asleep); and (iii) musculoskeletal (i.e. joint or muscle pain).
Participation was significantly higher among females than malesOR was adjusted for age, gender, rank, service branch, component, income and education level.Overall risk of bias: Low
Unwin, 1999 & 2002(9, 88)
Cross-sectional postal survey, 1997
UK male and female veterans of the 1991 Gulf War; 1992-1997 Bosnia war and non-deployed veterans (also termed Gulf-Era veterans).N= 3,510 GWVN= 2,040 Bosnia veteransN= 2,600 Gulf era veterans
CDC-derived definition: Presence, for a month or more, of at least one symptom from two or more symptom groupings namely (i) fatigue; (ii) mood/cognition (i.e. depression, poor concentration/memory, moodiness, anxiety, word-finding difficulties or sleep difficulties); and (iii) musculoskeletal (i.e. joint or muscle pain, joint stiffness)
62.0 Bosnia: 36.6Gulf-Era: 36.3
Bosnia: 2.42Gulf-Era: 2.41
2.06-2.85 †2.10-2.76 †
Participation rates: 70% in GWV, 62% in Bosnia veterans and 63% in Gulf-Era group.Non-response bias was assessed and sex distribution and number of medical discharges were similar for participants and non-participants. Participants were more likely to be older or still servingOR was adjusted for age, marital status, rank, education, employment, discharge status, smoking status, alcohol consumption, general health questionnaire scoresOverall risk of bias: Low
Kelsall, 2009 (20)
Cross-sectional postal survey, 2000-2002
Australian male veterans of the 1991 Gulf War and non-Gulf War deployed veterans .N= 1,381 GWVN= 1,377 non-GWV
CDC-modified definition: Presence, for a month or more, of at least one symptom rated as moderate/severe from three or more symptom groupings namely (i) fatigue; (ii) psycho-physiological distress (e.g. vomiting/nausea, stomach cramps, diarrhoea, wheezing, indigestion, persistent cough, fainting, dizziness, difficulty speaking); (iii) cognitive distress (e.g. loss of concentration,
25.6 16.0 1.80 1.48-2.19 Participation rates: 81% in GWV and 57% in non-GWV.Non-response bias was assessed and participants were more likely to be older and of higher ranks than non-participants. Participation of Air Force personnel was higher among non-GWV than GWV.ORs were adjusted for age, service branch, rank, marital
115
First Author
Study design; study period
Sample Multisymptom illness (MSI)case definition*
GWV prevalence
(%)
Comp group prevalence
(%)
ES(Adj. OR)
95% CI Comments and Assessment of overall risk of bias
feeling distant, sleeping difficulties, distressing dreams, irritability/outbursts of anger); and (iv) arthro-neuromuscular distress (e.g. joint stiffness, general muscle aches, low back pain, joint pain without swelling or redness). The symptom groupings psycho-physiological distress, cognitive distress and arthro-neuromuscular distress were derived through exploratory factor analysis of symptoms reported by Australian GWV (108)
status and highest level of educationOverall risk of bias: Low
Blanchard, 2006 (21)
Cross-sectional face-to-face interviews, 2001
US Gulf War veterans and non-deployed veterans.Ν=1,035 GWVΝ= 1,116 non-GWV
CDC-derived definition: Presence, for 6 months or more, of at least one symptom from two or more symptom groupings namely (i) general fatigue; (ii) mood/cognition (i.e. feeling depressed, feeling irritable, difficulty thinking/concentrating, feeling worried/tense/anxious, problems finding words, problems getting to sleep); and (iii) musculoskeletal (i.e. joint pain, muscle aches/pain)
28.9 15.8 2.16 1.61-2.90 Participation rates: 53% in GWV & 39% in non-GWVNon-response bias was assessed and participants were more likely to be female, older, white, in the reserve and using results of an earlier study were more likely to have reported more symptoms and illnesses than non-participantsOR was adjusted for age, gender, race, education, duty type, service branch, rank, income, combat exposure, Khamisiyah exposure, psychiatric conditions with onset prior January 1 1991 and self-reported doctor diagnosed medical conditionsOverall risk of bias: Low
Kang, 2009 (31)
Cross sectional postal and telephone
US GWV and non-deployed Gulf Era personnel. N = 6,111 GWV
Presence of several different symptoms together that persisted for 6 months or more and could not be adequately explained through medical or
36.5 11.7 3.05 2.77-3.36 Participation rates: 40% GWV and 27% non-GWV.Non-response bias was assessed and non-respondents were more
116
First Author
Study design; study period
Sample Multisymptom illness (MSI)case definition*
GWV prevalence
(%)
Comp group prevalence
(%)
ES(Adj. OR)
95% CI Comments and Assessment of overall risk of bias
interviews, 2003-2005
N = 3,859 non-GWV psychiatric diagnoses. The list of symptoms might include things like fatigue, muscle/joint pain, headaches, memory problems, digestive problems
likely to be younger, single, non-white or enlisted rank in 1991 than participantsDefinition of multisymptom illness adopted in the study was similar to CDC but not clearly definedOR was adjusted for age, gender, race, body mass index, cigarette smoking, rank, service branch, unit component (active duty, National Guard or reserve)Overall risk of bias: High
Abbreviations: GWV=Gulf War veterans; non-GWV=non-Gulf War veterans; ES=effect size; OR=odds ratio; CI=confidence interval
* The CDC-derived definitions are multisymptom illness definitions which were developed on the basis of the multisymptom illness definition developed by Fukuda et al. (1998)(27), which was also referred to as the CDC definition. Variations in symptoms listed under each of the three symptom groupings depended on the list of symptoms used in each of the studies.
† The odds ratios were combined for males and females
117
5.2.2 Key findings
Seven eligible articles were identified reporting multisymptom illness according to the eligibility
criteria which included studies that used the CDC definition for multisymptom illness. All of these
studies were in Gulf War veteran populations, there were no eligible studies identified in
Afghanistan or Iraq War veteran populations. Meta-analysis of the seven eligible studies showed
that the odds of multisymptom illness were more than two and a half times greater in Gulf War
veterans compared with non-deployed comparison groups. There was a high amount of statistical
heterogeneity. Subanalysis conducted based on five studies that used the CDC definition of three
groupings of symptoms to define a multisymptom illness case showed that the OR increased
slightly to around three fold in Gulf War veterans compared with non-deployed military comparison
group. The odds of multisymptom illness and the heterogeneity in studies decreased in the
analysis involving just the four studies assessed as overall low risk of bias.
118
6 Chronic fatigue syndrome in Gulf War, Afghanistan and Iraq War veterans compared to non-deployed military personnel
6.1 Literature search results
Figure 34 shows that the search yielded 1,721 records, with 1,332 records remaining after removal
of duplicates. The titles and abstracts were screened to identify studies for full- text review by the
specified inclusion and exclusion criteria. After abstract review, 71 full-text articles were identified
for further review, 11 eligible articles were identified reporting CFS according to eligibility criteria.
The reasons for excluding full text articles assessed for eligibility for CFS are also identified.
119
1721 Records Identified Through Database Search
405 Duplicates Removed
1332 Records After Duplicates Removed
1332 abstracts assessed for eligibility
1261 abstracts excluded
71 full-text articles assessed for eligibility
60 full-text articles excluded due to:
3 articles – non-specified deployment or non-Gulf/ Afghanistan/Iraq War deployment
28 articles – inappropriate or no comparison group or case definition of CFS
15 articles – fatigue outcome but no assessment of CFS
13 articles – not original research or odds ratio not calculable
1 article - treatment seeking sample
11 Eligible Articles Reporting CFS 11 in Gulf War veterans
0 in Afghanistan/Iraq War veterans
7 studies included in quantitative synthesis reporting CFS in Gulf
War veterans4 eligible articles excluded due to
overlap in study populations
16 Records Identified through Other Sources, e.g. reference lists, grey
literature
Figure 34 PRISMA flow diagram for chronic fatigue syndrome (CFS) in Gulf War, Afghanistan, Iraq War veterans
120
6.2 CFS in Gulf War veterans compared to non-deployed military personnel
6.2.1 Results
Seven studies met the eligibility criteria (19, 29, 31, 83, 106, 109, 110). The sample of veterans in
Kang 2003 and Kang 2009 overlapped so Kang 2009 (31) was excluded from the meta-analysis.
Proctor 2001 (106) reported prevalence data only and the prevalence was zero in the non-
deployed group, so an unadjusted OR was calculated by adding 0.5 to all cells.
Based on a random-effects meta-analysis, there was an increased odds of CFS in deployed
compared with non-deployed veterans (OR 7.62, 95% CI 3.91 to 14.85) (Figure 35). There was a
moderate amount of statistical heterogeneity (I2 = 52%).
Figure 35 Random-effects meta-analysis illustrating log-transformed odds ratios of chronic fatigue syndrome in Gulf War veterans and non-deployed military personnel
In a sensitivity analysis excluding one study which was rated at high risk of bias and which
reported an unadjusted OR (Proctor 2001) (106), the meta-analytic effect increased to OR 8.21
(95% CI 3.99 to 16.89), but statistical heterogeneity remained high (I2 = 61%). In sensitivity
121
analyses excluding each study one by one, all meta-analytic ORs remained the same magnitude,
except when Eisen 2005 (109) was removed (Table 1). There were not enough studies to perform
subgroup analyses or to assess for small study effects.
Table 14 Sensitivity analyses excluding each study one by one for studies of CFS in Gulf War veterans compared to non-deployed military personnel
Excluded study Pooled OR LCI 95% HCI 95% Cochran Q Chi2 P-value I2 (%)
Eisen 2005 4.94 4.05 6.02 1.67 0.80 0
Ismail 2008 7.84 3.35 18.33 9.93 0.04 60
Kang 2003 9.97 4.49 22.12 5.67 0.23 29
Kelsall 2006 8.15 3.78 17.59 10.45 0.03 62
Proctor 2001 8.21 3.99 16.89 10.17 0.04 61
Steele 2000 7.73 3.34 17.91 9.81 0.04 59
In the Kang 2009 study (31) which was excluded from the meta-analysis, the authors reported
prevalence data and adjusted risk ratios (RR), so we calculated unadjusted ORs based on the
prevalence data. There was an increased odds of CFS in deployed compared with non-deployed
veterans (OR 2.93, 95% CI 2.4 to 3.58; note that the adjusted RR in the paper was RR 2.38, 95%
CI 1.97 to 2.87).
6.2.2 Key findings
Seven studies met the eligibility criteria for the systematic review and six studies for inclusion in the
meta-analysis of CFS in Gulf War veterans. The number of cases of CFS in some study groups,
and non-deployed comparison groups in particular was small or none. Based on a random-effects
meta-analysis, there was an increased odds of CFS in Gulf War veterans compared with non-
deployed military personnel of over seven-fold. There was a moderate amount of statistical
heterogeneity. In sensitivity analyses, excluding each study one by one, all meta-analytic ORs
remained the same magnitude, except when Eisen 2005 (109) was removed and the OR
decreased to around four-fold of CFS in Gulf War veterans compared with non-deployed military
personnel. There were no reported studies of CFS in Afghanistan/Iraq War veterans.
122
Table 15 Characteristics of eligible studies comparing prevalence of chronic fatigue syndrome (CFS) and CFS-like illness in Gulf War veterans and non-deployed military personnel
First Author
Study design; study period
Sample CFS definition
Outcome GWV prevalence
(%)
Comp group prevalence
(%)
Effect measure 95% Confidence Interval
Participation rates, Assessment of overall risk of bias
Steele (2000) (29)
Cross-sectional telephone interview; 1998
Kansas residents who had served active military duty any period between August 1990 and July 1991 and now separated or reserve Stratified by gender and reservist status
Cases defined on basis of self-reported symptoms, fatigue characteristics and medical diagnoses, according to 1994 definition. Also self-reported CFS diagnosed by a physician
Self-reported CFS diagnosed or treated by a physician
9.0 1.0 OR 8.7 3.53, 21.46Participation rates: 93% GWV; 73% non-GWV, 91% of males, 95% of females completed the interviews of all GWV and non-GWVFor self-reported CFS diagnosed or treated by a physician, OR adjusted for sex, age, income, education level. For CFS-like illness, OR also adjusted for rank, service branch and duty statusOverall risk of bias: low
CFS-like illness 7.1 0.7 OR 8.21 2.58, 26.10
Proctor (2001) (106)
One phase of a longitudinal study of the Devens Cohort; 1994-1996
Stratified random sample (high and low symptom reporters) of the original Devens Cohort and German deployed comparison group
1994 definition based on self-reported symptoms, categorised before and after exclusionary indications ruled out
CFS-like illness before exclusionary indications ruled out
7.5 0 Not estimable Participation rates: Not reportedNon-response bias not assessedOverall risk of bias: high
CFS-like illness after exclusionary indications ruled out
2.0 0 Not estimable
Kang (2003) (83)
Phase 1 of the longitudinal National Health Survey of Gulf War era veterans and
Stratified random sample of US Gulf War veterans (Females=3000, NG=4000,
CFS-like illness defined by the 1994 definition based on self-report
CFS-like illness 5.6 1.2
Unadjusted OR
5.0 4.1, 6.2 Participation rate: 70% overall (76.3% GWV vs 63.2% non-GWV) eligible in-state and contactableNon-response bias
Adjusted OR 4.8 3.9, 5.9
123
First Author
Study design; study period
Sample CFS definition
Outcome GWV prevalence
(%)
Comp group prevalence
(%)
Effect measure 95% Confidence Interval
Participation rates, Assessment of overall risk of bias
their families; 1995-1996
Res=5000) symptoms assessed for demographics; significant differences between responders and non-responders on age, race, marital status and rankOR adjusted for age, marital status, rank and unit componentOverall risk of bias: low
Eisen (2005) (109)
Phase 2 of the longitudinal National Health Survey of Gulf War era veterans and their Families; 1999-2001
All eligible Gulf War veterans and comparison group members from the 1995 sample
1994 definition of CFS and self-reported CFS
1994 defined CFS 1.6 0.1
Unadjusted OR
17.68 4.63, 67.57 Participation rates: 53.1% (1061 of the 1996 invited to participate) GWV; 39.1% non-deployed (1128 of the 2883 invited to participate)Participation bias assessed for demographics, health outcomes: significant differences between responders and non-responders found for demographics and some service characteristics. Differences also found for some symptoms, including those suggestive of CFSFor both outcomes ORs were adjusted for age, sex, race, smoking, duty type, service branch and rank (for self-reported CFS the OR was also
Adjusted OR 40.60 10.2, 161.15
Self-reported CFS
2.3 0.4 Unadjusted OR
5.61 1.71, 18.42
Adjusted OR 8.05 1.94, 33.43
124
First Author
Study design; study period
Sample CFS definition
Outcome GWV prevalence
(%)
Comp group prevalence
(%)
Effect measure 95% Confidence Interval
Participation rates, Assessment of overall risk of biasadjusted for education)Overall risk of bias: low
Kang (2009) (31)
Phase 3 of the longitudinal National Health Survey of Gulf War era veterans and their Families; 2003-2005
All eligible veterans from the original study populations (Kang 2003) were invited to participate
CFS-like illness defined by the 1994 definition based on self-report symptoms
CFS-like illness
9.4 3.4 Unadjusted OR
2.93 2.40, 3.58 Participation rate: 34%. Gulf War veterans (6111; 80% males) and Gulf War-era comparison group (3589; 78% males)Non-response bias assessed for demographics; significant differences between responders and non-responders on age, race, marital status and rankRisk ratio adjusted for age, gender, race, rank, unit component, branch of service, BMI and current smoking historyOverall risk of bias: low
Kelsall (2006) (19)
Cross-sectional medical interview, structured diagnostic interview administered by trained psychologists and self-report questionnaire; 2000-2002.
All Australian GWV (n = 1871) and a random sample of non-deployed era personnel (n = 2796)
1994 definition of CFS and medically unexplained chronic fatigue
CFS 0.8 0.1 Adjusted OR 5.1 1.1, 48.5 Participation rates: 80.5% GWV (n=1456); 56.8% era non-deployed (n=1588).Non-response bias assessed for demographics; significant differences for age, service branch, rank and education.For CFS, OR adjusted for age, service branch and rank. For other outcomes, OR adjusted for age, service branch,
Medically unexplained chronic fatigue
6.6 2.9 Adjusted OR 2.3 1.6, 3.4
125
First Author
Study design; study period
Sample CFS definition
Outcome GWV prevalence
(%)
Comp group prevalence
(%)
Effect measure 95% Confidence Interval
Participation rates, Assessment of overall risk of biasrank, education, marital status, smoking and alcohol useOverall risk of bias: low
Idiopathic chronic fatigue
5.7 2.9 Adjusted OR 2.1 1.4, 3.2
Ismail (2008) (110)
Phase 2 of a longitudinal UK veteran study; postal survey at phase 1 (1997-1998) and medical assessment at phase 2 (1999-2000)
At phase 1, population based survey UK Armed Forces Gulf War era veterans. Random sample of Gulf War veterans, deployed controls (Bosnia) and non-deployed controls. At phase 2, random sample who screened positive for physical disability at phase 1
1994 definition of CFS
CFS 18.0 3.0 Unadjusted OR
10.6 3.4, 32.9 Participation rate: overall 29.7% at phase 2Non-response bias not assessed. Comparison between disabled-GWV and disabled non-GW personnel indicated significant differences in GWV more likely to have discharged and differences in rankEffect estimate adjusted for age, sex, rank, marital status, alcohol-related disorders and selection bias using probability weightsOverall risk of bias: low
CFS Adjusted OR 7.8 2.5, 24.5
CFS with DSM-IV depression or anxiety
10.8 1.5 Unadjusted OR
7.9 1.7, 36.3
CFS with DSM-IV depression or anxiety
Adjusted OR 9.1 1.8, 47.0
126
7 DiscussionThe overall pattern of findings from these systematic reviews of psychological disorders,
multisymptom illness and CFS in Gulf War veterans, Afghanistan and Iraq War veterans was that
for virtually all of the psychological disorders, the meta-analytical summary ORs were elevated in
Gulf War or Afghanistan/Iraq War veterans compared with military personnel who were not
deployed to the corresponding conflict zone. The summary ORs for PTSD and GAD in Gulf War
veterans compared with non-deployed personnel were statistically significantly higher than the
summary ORs of PTSD and GAD in Afghanistan/Iraq War veterans respectively. The summary
odds of depression were higher in Gulf War veterans than in Afghanistan/Iraq War veterans and
this narrowly missed statistical significance. Substance use disorders were relatively under
researched, with a small number of studies having been identified through the systematic review.
The meta-analytical ORs of multisymptom illness and of CFS in 1991 Gulf War veterans compared
with non-deployed personnel were elevated. There were no studies identified through the search
strategy for the systematic review of multisymptom illness and of CFS in Afghanistan and Iraq War
veterans.
Table 16 summarises the main meta-analysis summary ORs for psychological disorders,
multisymptom illness and CFS, and the assessment of difference between the summary ORs in
Gulf War veterans and in Afghanistan/Iraq War veterans where applicable.
Table 16 Main meta-analysis summary odds ratios for psychological disorders, multisymptom illness and chronic fatigue syndrome (CFS) in Gulf War, Afghanistan/Iraq War veterans compared with non-deployed personnel
Main health outcome Gulf War veterans
OR (95% CI). I2
Afghanistan/Iraq War veterans
OR (95% CI). I2
P-value for the test for equality of the summary OR
Depression 2.28 (1.88-2.76). I2=75% 1.58 (1.14-2.17). I2=98% 0.055
Dysthymia or chronic dysphoria
2.39 (2.0-2.86). I2=0% n.a. n.a.
PTSD 3.39 (2.79-4.13). I2=53% 2.12 (1.65-2.72). I2=97% 0.004
Alcohol use disorders 1.33 (1.22-1.46). I2=14% 1.36 (1.11-1.66). I2=77% 0.862
Substance use disorders 2.13 (0.96-4.72). I2=29% 1.14 (1.04-1.25). I2=0% 0.053
Any substance use disorder 1.35 (1.25-1.46). I2=0% * n.a.
Generalised anxiety disorder 3.04 (1.95-4.75). I2=35% 1.20 (1.00-1.44). I2=0% Respective 95% CI did not overlap
Multisymptom illness 2.74 (2.15-3.51). I2=92% † n.a.
CFS 7.62 (3.91-14.85). I2 = 52% † n.a.
n.a. Not applicable.* ‘Any substance use’ outcome was measured in Afghanistan/Iraq War veterans’ study (78) and was reported separately.
127
† There were no studies identified of multisymptom illness or CFS in Afghanistan/Iraq War veterans.
The mental health in reservists compared with active duty/regular veteran personnel post
deployment has been of concern in relation to the increased proportion of US or UK National
Guard or reservist forces deployed to Afghanistan/Iraq (111-114), and although the relationship
between active or reservist status on psychological health post deployment was not a primary
research question in this project, regular versus reservist status was a subgroup analysis
undertaken for a limited number of psychological outcomes. There were no important differences
in the likelihood of PTSD in reservists compared with regular personnel deployed to the Gulf War
or to Afghanistan/Iraq War. The summary estimate of the likelihood of alcohol use disorders was
slightly higher, but not statistically significant, in reservists compared with regular personnel
deployed to Afghanistan/Iraq War.
This series of systematic reviews and meta-analyses of international literature had several
strengths. Studies were identified through a search using multiple electronic databases from 1
January 1990 to 24 January 2014. Two study team members independently reviewed abstracts
and assessed possible papers for eligibility. We assessed the overall risk of bias in individual
studies that were included in the systematic reviews and meta-analyses, which was not performed
in other reviews (11, 15, 24, 26) of a similar nature, and the robustness of the results to risk of bias
through subanalysis. However, many of these previous reviews discussed essential biases of
included studies individually, and a narrative description of the studies, but this is not as strong as
a formal assessment of risk of bias.
Our rigorous methodology of only including studies with a military non-deployed comparison group
eliminated many methodological concerns of the previous reviews in the field by concentrating on
the psychological outcome or multisymptom illness or CFS as the condition of interest, and which
may have included papers which used civilian comparison groups. We excluded treatment
seeking self-selected populations that are more likely to experience higher rates of disorders and
may not be representative of the overall military population that were deployed. We also excluded
several well conducted prevalence studies that did not have any military comparison group, since
comparing prevalence studies using different methodologies makes it very difficult to ascertain
whether the differences in associations were due to difference in deployment or different
methodologies used in the studies rather than comparison with similar but non-deployed military
personnel (37). A limitation was the small number of studies identified in some reviews, such as
reported substance use disorders in Gulf War and Afghanistan/Iraq War veterans, which limited the
statistical power in some meta-analyses that involved a small number of studies and also limited
the capacity to undertake further subgroup analyses for some outcomes. However, given our
extensive search methodology, we consider it unlikely that we missed relevant literature.
128
Our systematic reviews and meta-analyses included studies across veteran cohorts from the US,
UK, Australia, Germany and Canada, the majority of which sampled across all three services. The
deployed forces of the countries from which the veteran cohorts were selected varied, but the
included studies all included a non-deployed comparison group and the majority adjusted for
possible demographic and military service confounding factors. The main aim in these systematic
reviews was to consider the summary ORs comparing the health outcome of interest in deployed
compared with non-deployed personnel combining the ORs from the eligible studies.
Although PTSD has often received more attention and publicity in relation to military and veterans’
health, we are not aware of any previous systematic reviews that have estimated comprehensively
the risk of depression, alcohol use disorders, substance use disorders, and GAD in Gulf War and
Afghanistan/Iraq War veterans. Our systematic reviews have extended and updated previous
reviews undertaken for multisymptom illness and chronic fatigue and symptom based conditions
and have also undertaken meta-analyses for the first time.
Multisymptom illness and CFS have been identified contextually more with the 1990-1991 Gulf War
than with deployments to Afghanistan and the Iraq War. However, to the knowledge of the authors
of this report, a comprehensive systematic review of the literature on these symptom based
conditions in veterans of other deployments to this area of operations, has not been undertaken.
The eligibility criteria in our systematic review for inclusion of studies regarding multisymptom
illness and CFS included internationally accepted case definitions for the epidemiological study of
these conditions (19, 20, 27, 35). The US IOM 2014 report on development of a consensus
definition on chronic multisymptom illness (115) was published towards the end of the research
process, but the IOM Committee recommended that the US Veterans Affairs (VA) consider the use
of the CDC definition (and Kansas definition (29)) of multisymptom illness because they capture
the most commonly reported symptoms by Gulf War veterans.
Gulf War veteran health studies used a variety of definitions to investigate fatigue related outcomes
in veterans including chronic fatigue caseness based on the Chalder Fatigue Scale (116), however
we used the more rigorous 1994 international definition (19, 35) for defining CFS as the outcome
measure for inclusion. It is possible that studies of Afghanistan and Iraq veterans have assessed
fatigue in veterans based on less rigorous definitions than CFS, including screening instruments
and scales, but these were not considered in this review. There are also various other approaches
to defining multisymptom illnesses in Gulf War veterans, including defining Gulf War illnesses, that
were not included in our review, based on our eligibility criteria.
The focus of each of the systematic reviews was the outcome under consideration, eg depression
or PTSD or alcohol use disorder. We recognise that many Gulf War veterans and comparison
group subjects with depression, for example, may also meet criteria for other psychological
129
disorders including PTSD, substance use disorders, and anxiety disorders (1, 117). A detailed
examination of this comorbidity between psychological disorders, however, was beyond the scope
of this review and would detract from the primary focus.
Undertaking these systematic reviews has assisted in drawing conclusions about consistency of
the results of studies in relation to psychological outcomes in Gulf War veterans, Afghanistan and
Iraq War veterans, and multisymptom illness and CFS in Gulf War veterans compared to personnel
who were not deployed to a war zone or who were deployed elsewhere. Conducting meta-
analyses and presenting the outputs produced visual and comparable summary effect estimates of
these outcomes in Gulf War veterans, Afghanistan and Iraq War veterans compared with non-
deployed military personnel and quantified this in overall summary measures. By reporting
summary estimates, it is easier and quicker for readers of this report, including health policy
makers and service providers, non-researcher veterans, and clinicians to gain an overview of the
relevant literature.
8 Implications of findingsAn important finding from the current systematic reviews and meta-analyses is that virtually all of
the psychological disorders under study of depression, PTSD, alcohol use disorder and substance
use disorders, and GAD were elevated in troops deployed to the Middle East area of conflicts over
the past 20 years compared with non-deployed military personnel. Although much attention and
general awareness has focused on PTSD and the increased risk of PTSD associated with
deployment to war and conflict zones, substantially less has focused on depression and other
psychological disorders. These systematic reviews and meta-analyses show that these
psychological conditions are also elevated in Gulf War and Afghanistan/Iraq War deployed
compared with non-deployed personnel and poorer psychological health is not restricted to PTSD.
A further important finding is that the odds of multisymptom illness and CFS were significantly
elevated in Gulf War veterans compared with non-deployed military personnel. Studies meeting
the inclusion criteria for the systematic review of multisymptom illness and of CFS were identified
in Gulf War veteran study populations, but not in Afghanistan or Iraq War veteran study
populations. This could suggest that these conditions were not a primary concern or complaint in
the context of Afghanistan/Iraq War veterans’ health. It is somewhat surprising that studies of
Afghanistan/Iraq War veterans did not include a case definition of multisymptom illness or CFS
similar to that employed in studies of previous veterans to that conflict zone, however they may
have employed alternative or less rigorous definitions that did not meet our inclusion criteria.
130
The most likely reason for the finding of increased alcohol use disorders and substance use
disorders is that alcohol and other drugs are being used to “self-medicate” – to ameliorate other
psychological or physical problems (118). This seems a plausible explanation given the high rates
of depression and other psychological conditions in troops deployed to the Gulf War (11) and to the
Afghanistan/Iraq War (73). Although the small numbers of studies of substance use disorders in
Gulf War veterans limited power to detect a statistical difference between the study groups, stigma
around illicit substances may have resulted in some underreporting of other substance abuse. The
military handles alcohol and tobacco, legal substances, very differently to illicit substances.
Based on random-effects meta-analysis in the eligible studies, there was an increased odds of
depression and of PTSD in deployed Afghanistan/Iraq War veterans compared with non-deployed
personnel. However, there was a very high amount of statistical heterogeneity in each of the main
analyses, so the actual size of the increased meta-analytic effect should be interpreted with
caution. No explanation, investigated through subanalyses, was found for this high heterogeneity.
Possible explanations for consideration are that 1990-1991 Gulf War veterans who are now older
have more chronic conditions, which have stabilised, whereas psychological conditions in younger
veterans of more recent conflicts fluctuate more. It is well established that there is a dose
response relationship between combat exposure and PTSD, an underlying factor in the
heterogeneity observed in the meta-analysis of the association between PTSD and
Afghanistan/Iraq War deployment compared with that in the meta-analysis of PTSD and Gulf War
deployment may be a greater variation in the types of experiences veterans had in the more recent
conflicts in the Middle East. Other Afghanistan/Iraq War deployment related factors such as the
possibility of multiple deployments, variables within deployments, the chronicity of the overall
deployment period, and the potential for other deployments at a time of increased operational
tempo may also have contributed to this observed heterogeneity.
The odds of psychological disorders were all slightly greater in Gulf War veterans and the
summary ORs for PTSD and GAD were statistically significantly higher than in Afghanistan/Iraq
War veterans, and may reflect a level of chronicity. Psychological disorders tend to increase post
deployment; the rate of onset of symptomatology of each broad DSM-IV diagnostic category was
found to peak in the first two years following the Gulf War, in Australian Gulf War veterans, and
then subsided. This pattern was particularly noticeable in the case of alcohol disorders (117).
The presentation of idiopathic physical symptoms including multisymptom illness and fatigue
related conditions including symptoms of fatigue and chronic fatigue are likely to be to the person’s
general medical practitioner. The presentation of these symptoms and conditions are not as likely
to be through the mental health system. With overall increased odds of multisymptom illness and
of CFS in Gulf War veterans, departments of veterans’ affairs (Australian and overseas) need to
131
work with general practitioners and primary care providers in relation to this symptomatology and
symptom burden and conditions.
Our findings have important policy and program implications. About 697,000 US troops were
deployed to the 1991 Gulf War, with other coalition forces (from countries such as the UK, France,
Canada and Australia) amounting to nearly 260,000 at their peak personnel strength (119). Over
two million US veterans have deployed to the Afghanistan and Iraq conflicts (120) in a coalition of
49 countries with the UK providing the second largest force. Increased risk of any condition in
veteran populations of these sizes is clearly a concern.
While disorders such as PTSD and depression have been the primary focus, it is also important
that the elevated risk of substance use disorders in veterans is recognised, as there is a strong
association between those disorders and substance use disorders (121, 122). Furthermore,
individuals with PTSD and depression co-occurring with substance use disorders often have worse
treatment outcomes (123). Substance (and alcohol) use disorders are particularly troubling
because of the powerful impact on behaviour, on the individual’s health as well as impacts on
family, community and society as a whole, in addition to the difficulties in diagnosis and
management of comorbid disorders (124).
Further, there are circumstances and exposures associated with the Afghanistan and Iraq War
deployments which may render a systematic review of the literature thus far highly pertinent. For
example, the level of traumatic brain injury from these deployments has been associated with
increased psychiatric illness outcomes (125). An examination of the relationship between
traumatic brain injury and psychological disorders was not in the scope of this review.
An investigation by Hoge (73) reported that combat duty in Iraq was associated with high utilisation
of mental health services and attrition from military service after deployment, and that the high rate
of mental health services utilisation post-deployment is a challenge for resource allocation. In the
context of our reviews which have found the summary OR for the odds of all psychological
disorders to be elevated in Afghanistan/Iraq War veterans and virtually all in Gulf War veterans this
finding has important implications for health service policy and delivery.
9 Implications for future researchOur systematic review showed that substance use disorders were generally under researched. For
example, further studies with increased power are needed to assess substance use disorder risk in
Gulf War veteran populations, and this is a consideration for future research studies.
132
Our systematic reviews considered psychological health outcomes, multisymptom illness and CFS
and considered possible explanations for heterogeneity across studies. The methodology could be
extended to assess the effect of important risk factors for these health outcomes, including military
service exposures, different types of forces deployed e.g. branch of service, and vulnerable
populations e.g. gender differences in deployed personnel. The differences that were observed in
the summary ORs in health outcomes between Gulf War veterans and Afghanistan/Iraq War
veterans could be compared again in the future, repeating the search and systematic review to
update the studies and the meta-analyses, to investigate whether the differences change over
time.
Influences on health outcomes are multifactorial but the influence of cultural factors and veteran
and general population health care systems and services and uptake of those services for the
health outcomes of interest could provide a useful comparison between countries that have
deployed forces to the same conflicts, identify potential gaps, and service as an evidence base for
future interventions.
10 ConclusionOur systematic reviews and meta-analyses found that depression, PTSD, alcohol use disorder and
GAD were elevated in Gulf War veterans and Afghanistan/Iraq War veterans compared with non-
deployed military personnel, with statistically significant higher risks found for Gulf War veterans’
summary odds in relation to PTSD and GAD compared with Afghanistan/Iraq veterans. Any
substance use disorders (i.e. alcohol and/or other substance use disorders) were elevated in Gulf
War veterans compared with non-deployed personnel, and substance use disorders were elevated
in Afghanistan/Iraq War veterans compared with non-deployed military personnel, but further
studies with increased statistical power are needed to assess the association with substance use
disorders in Gulf War veterans. Our systematic reviews and meta-analyses also found that the
odds of multisymptom illness and CFS were significantly elevated in Gulf War veterans compared
with non-deployed military personnel, and no studies of Afghanistan/Iraq War veterans met the
inclusion criteria for review of these disorders.
A further important finding is that the odds of multisymptom illness and CFS were significantly
elevated in Gulf War veterans compared with non-deployed military personnel. Studies meeting
our eligibility criteria were not identified in Afghanistan/Iraq War veterans. Idiopathic symptoms
contributing to multisymptom illness and CFS are more likely to present to general practitioners
than through the mental health system and Veterans’ Affairs departments could work with general
and primary care providers to address this burden of symptom reporting.
133
The main aim of these systematic reviews and meta-analyses was to consider the summary ORs
comparing the health outcome of interest in deployed compared with non-deployed personnel
combining the ORs from the eligible studies. However, the methodology could be extended to
assess the effect of important risk factors for these health outcomes, including different types of
forces deployed e.g. branch of service, military service exposures, and vulnerable populations e.g.
gender differences. This body of research has highlighted some key areas that warrant
consideration for policy makers and future researchers including cultural differences, the length of
time since the war and deployment and the impact on health, the influence of veteran and general
population health care systems between countries that are some of the factors that influence
health outcomes and could be compared as a basis for future policy development and intervention.
134
11 References1. Ikin JF, Sim MR, Creamer MC, et al. War-related psychological stressors and risk of
psychological disorders in Australian veterans of the 1991 Gulf War. Br J Psychiatry.
2004;185:116-26.
2. Australian Defence Force. The War in Iraq. Australian Operations in the Middle East in 2003.
2004. Available at: http://www.defence.gov.au/publications/lessons.pdf Accessed: 25 June,
2015
3. Fear NT, Jones M, Murphy D, et al. What are the consequences of deployment to Iraq and
Afghanistan on the mental health of the UK armed forces? A cohort study. Lancet.
2010;375:1783-97.
4. O’Bryant J, Waterhouse M. CRS Report for Congress. U.S. Forces in Iraq,. Washington
D.C.: Department of Defense; 2008. Available at:
http://www.fas.org/sgp/crs/mideast/RS22449.pdf. Accessed: 25 June, 2015.
5. Tanielian TL, Jaycox LH. Invisible wounds of war: psychological and cognitive injuries, their
consequences, and services to assist recovery. Santa Monica, California: RAND
Corporation; 2008.
6. Wolfe J, Proctor SP, Erickson DJ, et al. Relationship of psychiatric status to Gulf War
veterans' health problems. Psychosom Med. 1999;61:532-40.
7. Goss Gilroy Inc. Health study of Canadian Forces personnel involved in the 1991 conflict in
the Persian Gulf. Ottawa, Canada: 1998. Volume 1.
8. Iowa Persian Gulf Study G. Self-reported illness and health status among Gulf War veterans.
A population-based study. JAMA. 1997;277:238-45.
9. Unwin C, Blatchley N, Coker W, et al. Health of UK servicemen who served in Persian Gulf
War. Lancet. 1999;353:169-78.
10. Brailey K, Vasterling JJ, Sutker PB. Psychological aftermath of participation in the Persian
Gulf War. Lundberg, Ante [Ed]. 1998:83-101.
11. Stimpson NJ, Thomas HV, Weightman AL, et al. Psychiatric disorder in veterans of the
Persian Gulf War of 1991. Systematic review. Br J Psychiatry. 2003;182:391-403.
135
12. Bentel DJ, Smith DE. Drug abuse in combat: The crisis of drugs and addiction among
American troops in Vietnam. J Psychoactive Drugs. 1971;4:23-33.
13. Micklewright S. The management of alcohol abuse in the Royal Navy. J R Nav Med Serv.
1993;79:29-31.
14. Moore MM. Alcoholism in military service. J Lab Clin Med. 1943;28:515-24.
15. Thomas HV, Stimpson NJ, Weightman AL, et al. Systematic review of multi-symptom
conditions in Gulf War veterans. Psychol Med. 2006;36:735-47.
16. McKenzie DP, Ikin JF, McFarlane AC, et al. Psychological health of Australian veterans of
the 1991 Gulf War: an assessment using the SF-12, GHQ-12 and PCL-S. Psychol Med.
2004;34:1419.
17. Black DW, Carney CP, Peloso PM, et al. Gulf War veterans with anxiety. Epidemiology.
2004;15:135-42.
18. Black DW, Carney CP, Forman-Hoffman VL, et al. Depression in veterans of the first Gulf
War and comparable military controls. Ann Clin Psychiatry. 2004;16:53-61.
19. Kelsall H, Sim M, McKenzie D, et al. Medically evaluated psychological and physical health
of Australian Gulf War veterans with chronic fatigue. J Psychosom Res. 2006;60:575-84.
20. Kelsall HL, McKenzie DP, Sim MR, et al. Physical, psychological, and functional
comorbidities of multisymptom illness in Australian male veterans of the 1991 Gulf War. Am J
Epidemiol. 2009;170:1048-56.
21. Blanchard MS, Eisen SA, Alpern R, et al. Chronic multisymptom illness complex in Gulf War I
veterans 10 years later. Am J Epidemiol. 2006;163:66-75.
22. Institute of Medicine. Gulf War and Health Volume 8 Update of health effects of serving in the
Gulf War. Washington, D.C.: The National Academies Press; 2010.
23. Institute of Medicine. Gulf War and Health, Volume 4: Health effects of serving in the Gulf
War. Washington, D.C.: The National Academies Press; 2006.
24. Kok BC, Herrell RK, Thomas JL, et al. Posttraumatic stress disorder associated with combat
service in Iraq or Afghanistan: reconciling prevalence differences between studies. J Nerv
Ment Dis. 2012;200:444-50.
136
25. Sundin J, Fear NT, Iversen A, et al. PTSD after deployment to Iraq: conflicting rates,
conflicting claims. Psychol Med. 2010;40:367-82.
26. Gadermann AM, Engel CC, Naifeh JA, et al. Prevalence of DSM-IV major depression among
U.S. military personnel: meta-analysis and simulation. Mil Med. 2012;177:47-59.
27. Fukuda K, Nisenbaum R, Stewart G, et al. Chronic multisymptom illness affecting Air Force
veterans of the Gulf War. JAMA. 1998;280:981-8.
28. Haley RW, Kurt TL, Hom J. Is there a Gulf War Syndrome? Searching for syndromes by
factor analysis of symptoms. JAMA. 1997;277:215-22.
29. Steele L. Prevalence and patterns of Gulf War illness in Kansas veterans: Association of
symptoms with characteristics of person, place, and time of military service. Am J Epidemiol.
2000;152:992-1002.
30. Bourdette DN, McCauley LA, Barkhuizen A, et al. Symptom factor analysis, clinical findings,
and functional status in a population-based case control study of Gulf War unexplained
illness. J Occup Environ Med. 2001;43:1026-40.
31. Kang HK, Li B, Mahan CM, et al. Health of US veterans of 1991 Gulf War: a follow-up survey
in 10 years. J Occup Environ Med. 2009;51:401-10.
32. Milanak ME, Gros DF, Magruder KM, et al. Prevalence and features of generalized anxiety
disorder in Department of Veteran Affairs primary care settings. Psychiatry Res.
2013;209:173-9.
33. American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders
Fourth Edition, Text Revision DSM-IV-TR. Washington, DC: American Psychiatric
Association; 2000.
34. American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders,
Fifth Edition Washington, DC: American Psychiatric Association; 2013.
35. Fukuda K, Straus SE, Hickie I, et al. The chronic fatigue syndrome: A comprehensive
approach to its definition and study. Ann Intern Med. 1994;121:953-59.
36. Magruder KM, Yeager DE. The prevalence of PTSD across war eras and the effect of
deployment on PTSD: A systematic review and meta-analysis. Psychiatric Annals.
2009;39:778-88.
137
37. Moher D, Liberati A, Tetzlaff J, et al. Preferred reporting items for systematic reviews and
meta-analyses: the PRISMA statement. PLoS Med. 2009;6:e1000097.
38. Robins LN, Wing J, Wittchen HU, et al. The Composite International Diagnostic Interview: an
epidemiologic instrument suitable for use in conjunction with different diagnostic systems and
in different cultures. Arch Gen Psychiatry. 1988;45:1069.
39. Gilbody S, Richards D, Brealey S, et al. Screening for depression in medical settings with the
Patient Health Questionnaire (PHQ): a diagnostic meta-analysis. J Gen Intern Med.
2007;22:1596-602.
40. Mitchell AJ, Vaze A, Rao S. Clinical diagnosis of depression in primary care: a meta-
analysis. Lancet. 2009;374:609-19.
41. Cochrane Handbook for Systematic Reviews of Interventions Version 5.1.0 [updated March
2011]. In: Higgins J, Altman DG, eds. The Cochrane Collaboration, 2011; 2011:187-241.
42. Hoy D, Brooks P, Woolf A, et al. Assessing risk of bias in prevalence studies: modification of
an existing tool and evidence of interrater agreement. J Clin Epidemiol. 2012;65:934-9.
43. Vos T, Flaxman AD, Naghavi M, et al. Years lived with disability (YLDs) for 1160 sequelae of
289 diseases and injuries 1990–2010: a systematic analysis for the Global Burden of
Disease Study 2010. Lancet. 2013;380:2163-96.
44. Cohen J. A coefficient of agreement for nominal scales. Educ Psychol Meas. 1960;20:37-46.
45. Deeks JJ, Higgins JPT on behalf of the Statistical Methods Group of The Cochrane
Collaboration Statistical algorithms in Review Manager 5, August 2010. In: Review Manager
(RevMan) 5.3. Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration;
2014.
46. DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials. 1986;7:177-88.
47. Deeks JJ, Altman DG, Bradburn MJ. Systematic Reviews in Health Care, Meta-Analysis in
Context In: Statistical methods for examining heterogeneity and combining results from
several studies in meta-analysis London, UK: BMJ Publishing Group; 2008:285-312.
48. McNeil D. Epidemiological research methods. New York: Wiley; 1996.
49. Higgins JPT, Thompson SG, Deeks JJ, et al. Measuring inconsistency in meta-analyses.
BMJ Case Rep. 2003;327:557-60.
138
50. Stevens JR, Nicholas G. Metahdep: meta-analysis of hierarchically dependent gene
expression studies. Bioinformatics. 2009;25:2619-20.
51. Stevens JR, Taylor AM. Hierarchical dependence in meta-analysis. JEBS. 2009;34:46-73.
52. Thompson SG, Higgins JPT. How should meta-regression analyses be undertaken and
interpreted? Stat Med. 2002;21:1559-74.
53. Higgins JPT, Green S. Cochrane Handbook for Systematic Reviews of Interventions
Chichester, UK: John Wiley & Sons; 2008.
54. Egger M, Davey Smith G, Schneider M, et al. Bias in meta-analysis detected by a simple,
graphical test. BMJ. 1997;315:629-34.
55. Sterne JA, Egger M. Funnel plots for detecting bias in meta-analysis: guidelines on choice of
axis. J Clin Epidemiol. 2001;54:1046-55.
56. Stata Statistical Software: Release 13. StataCorp. College Station, TX; 2011.
57. The Cochrane Collaboration. Glossary of terms in The Cochrane Collaboration [electronic
article]. 2005; Version 4.2.5.
58. Barendregt JJ, Doi SA. MetaXL version 1.4. [electronic article]. Available at:
http://www.epigear.com/index_files/metaxl.html Accessed 14 December, 2013.
59. Harbord RM, Higgins JPT. Meta-regression in Stata. Stata Journal 2008;8:493-519.
60. Perconte ST, Wilson AT, Pontius EB, et al. Psychological and war stress symptoms among
deployed and non-deployed reservists following the Persian Gulf War. Mil Med.
1993;158:516-21.
61. Sutker PB, Uddo M, Brailey K, et al. War-zone trauma and stress-related symptoms in
Operation Desert Shield/Storm (ODS) returnees. J Soc Issues. 1993;49:33-50.
62. Sutker PB, Davis JM, Uddo M, et al. Assessment of psychological distress in Persian Gulf
troops: ethnicity and gender comparisons. J Pers Assess. 1995;64:415-27.
63. Gray GC, Reed RJ, Kaiser KS, et al. Self-reported symptoms and medical conditions among
11,868 Gulf War-era veterans: The Seabee health study. Am J Epidemiol. 2002;155:1033-
44.
64. McCauley LA, Lasarev M, Sticker D, et al. Illness experience of Gulf War veterans possibly
exposed to chemical warfare agents. Am J Prev Med. 2002;23:200-6.
139
65. Ishoy T, Knop J, Suadicani P, et al. Increased psychological distress among Danish Gulf War
veterans--without evidence for a neurotoxic background. The Danish Gulf War Study. Dan
Med Bull. 2004;51:108-13.
66. Fiedler N, Ozakinci G, Hallman W, et al. Military deployment to the Gulf War as a risk factor
for psychiatric illness among US troops. Br J Psychiatry. 2006;188:453-9.
67. Toomey R, Kang HK, Karlinsky J, et al. Mental health of US Gulf War veterans 10 years after
the war. Br J Psychiatry. 2007;190:385-93.
68. Kang H, Mahan C, Lee K, et al. Illnesses among United States veterans of the Gulf War: a
population-based survey of 30,000 veterans. J Occup Environ Med. 2000;42:491.
69. Bray RM, Hourani LL, Rae Olmsted KL, et al. 2005 Department of Defense Survey of health
related behaviors among active duty military personnel. A component of the Defense
Lifestyle Assessment Program (DLAP). Prepared for the Assistant Secretary of Defense
(Health Affairs) under Cooperative Agreement No. DAMD17-00-2-0057 by RTI International,
Research Triangle Park, North Carolina; 2006. Available at:
(http://www.tricare.mil/hpae/_docs/2005%20Health%20Behaviors%20Survey_1-071.pdf).
Accessed 17 July, 2014.
70. Bray RM, Pemberton MR, Hourani LL, et al. 2008 Department of Defense Survey of health
related behaviors among active duty military personnel. A Component of the Defense
Lifestyle Assessment Program (DLAP),. Prepared for the Assistant Secretary of Defense
(Health Affairs) and for the U.S. Coast Guard, by RTI International, Research Triangle Park,
North Carolina; 2009. Available at: http://www.tricare/mil/tma/2008healthbehaviors.pdf.
Accessed: 29 July, 2014.
71. Barlas FM, Higgins WB, Pflieger JC, et al. 2011 Health Related Behaviors Survey of Active
Duty Military Personnel. Fairfax, Virginia: Department of Defense, TRICARE Management
Activity, Defense Health Cost Assessment and Program Evaluation (DHCAPE), and the
United States Coast Guard; 2013. Available at:
http://www.murray.senate.gov/public/_cache/files/889efd07-2475-40ee-b3b0-508947957a0f/
final-2011-hrb-active-duty-survey-report.pdf. Accessed: 17 July, 2014.
72. Cabrera OA, Hoge CW, Bliese PD, et al. Childhood adversity and combat as predictors of
depression and post-traumatic stress in deployed troops. Am J Prev Med. 2007;33:77-82.
140
73. Hoge CW, Auchterlonie JL, Milliken CS. Mental health problems, use of mental health
services, and attrition from military service after returning from deployment to Iraq or
Afghanistan. JAMA. 2006;295:1023-32.
74. Hoge CW, Castro CA, Messer SC, et al. Combat duty in Iraq and Afghanistan, mental health
problems, and barriers to care. N Engl J Med. 2004;351:13-22.
75. Kline A, Falca-Dodson M, Sussner B, et al. Effects of repeated deployment to Iraq and
Afghanistan on the health of New Jersey Army National Guard troops: implications for
military readiness. Am J Public Health. 2010;100:276-83.
76. Nguyen S, LeardMann CA, Smith B, et al. Is Military Deployment a Risk Factor for Maternal
Depression? J Womens Health Gend Based Med. 2013;22:9-18.
77. Peterson AL, Wong V, Haynes MF, et al. Documented combat-related mental health
problems in military noncombatants. J Trauma Stress. 2010;23:674-81.
78. Shen YC, Arkes J, Williams TV. Effects of Iraq/Afghanistan deployments on major
depression and substance use disorder: analysis of active duty personnel in the US military.
Am J Public Health. 2012;102 Suppl 1:S80-7.
79. Vanderploeg RD, Belanger HG, Horner RD, et al. Health outcomes associated with military
deployment: mild traumatic brain injury, blast, trauma, and combat associations in the Florida
National Guard. Arch Phys Med Rehabil. 2012;93:1887-95.
80. Vasterling JJ, Proctor SP, Amoroso P, et al. Neuropsychological outcomes of army
personnel following deployment to the Iraq war. J Am Med Assoc. 2006;296:519-29.
81. Wells TS, LeardMann CA, Fortuna SO, et al. A prospective study of depression following
combat deployment in support of the wars in Iraq and Afghanistan. Am J Public Health.
2010;100:90-9.
82. Holmes DT, Tariot PN, Cox C. Preliminary evidence of psychological distress among
reservists in the persian war. J Nerv Ment Dis. 1998;186:166-73.
83. Kang HK. Post-traumatic stress disorder and chronic fatigue syndrome-like illness among
Gulf War veterans: A population-based survey of 30,000 veterans. Am J Epidemiol.
2003;157:141-48.
84. Murphy D, Hooper R, French C, et al. Is the increased reporting of symptomatic ill health in
Gulf War veterans related to how one asks the question? J Psychosom Res. 2006;61:181-6.
141
85. Pontius EB, Wilson A, Perconte ST, et al. The need for continuing mental health intervention
in soldiers returning from the Persian Gulf War: assessment of deployed and non-deployed
reserve units from western Pennsylvania, eastern Ohio, and West Virginia. In: U. S.
Department of Veterans Affairs Persian Gulf Returnees Working Group, ed. Returning
Persian Gulf troops: first year findings: Northeast Program Evaluation Center; 1992:99-107.
86. Smith TC, Wingard DL, Ryan MAK, et al. PTSD prevalence, associated exposures, and
functional health outcomes in a large, population-based military cohort. Public Health Rep.
2009;124:90-102.
87. Stretch RH, Marlowe DH, Wright KM, et al. Post-traumatic stress disorder symptoms among
Gulf War veterans. Mil Med. 1996;161:407-10.
88. Unwin C, Hotopf M, Hull L, et al. Women in the Persian Gulf: Lack of gender differences in
long-term health effects of service in United Kingdom Armed Forces in the 1991 Persian Gulf
War. Mil Med. 2002;167:406-13.
89. Bleier J, McFarlane A, McGuire A, et al. Risk of adverse health outcomes associated with
frequency and duration of deployment with the Australian Defence Force. Mil Med.
2011;176:139-46.
90. Hotopf M, Hull L, Fear NT, et al. The health of UK military personnel who deployed to the
2003 Iraq war: a cohort study. Lancet. 2006;367:1731-41.
91. Hourani LL, Bray RM, Marsden ME, et al. 2006 Department of Defense Survey of Health
Related Behaviors Among the Guard and Reserve Force. A Component of the Defense
Lifestyle Assessment Program (DLAP). US Department of Defense; 2007. Accessed 1 July,
2014.
92. Jones M, Sundin J, Goodwin L, et al. What explains post-traumatic stress disorder (PTSD) in
UK service personnel: deployment or something else? Psychol Med. 2013;43:1703-12.
93. Shen YC, Arkes J, Pilgrim J. The effects of deployment intensity on post-traumatic stress
disorder: 2002-2006. Mil Med. 2009;174:217-23.
94. Shen YC, Arkes J, Kwan BW, et al. Effects of Iraq/Afghanistan deployments on PTSD
diagnoses for still active personnel in all four services. Mil Med. 2010;175:763-69.
95. Smith TC, Ryan MA, Wingard DL, et al. New onset and persistent symptoms of post-
traumatic stress disorder self reported after deployment and combat exposures: prospective
population based US military cohort study. BMJ. 2008;336:366-71.
142
96. Wittchen HU, Schonfeld S, Kirschbaum C, et al. Traumatic experiences and posttraumatic
stress disorder in soldiers following deployment abroad: how big is the hidden problem?
Dtsch Arztebl Int. 2012;109:559-68.
97. Vasterling JJ, Proctor SP, Friedman MJ, et al. PTSD symptom increases in Iraq-deployed
soldiers: comparison with nondeployed soldiers and associations with baseline symptoms,
deployment experiences, and postdeployment stress. J Trauma Stress. 2010;23:41-51.
98. Trautmann S, Schonfeld S, Behrendt S, et al. Substance use and substance use disorders in
recently deployed and never deployed soldiers. Drug Alcohol Depend. 2014;134:128-35.
99. Saunders JB, Aasland OG, Babor TFea. Development of the Alcohol Use Disorders
Identification Test (AUDIT): WHO Collaborative Project on Early Detection of Persons with
Harmful Alcohol Consumption--II. Addiction. 1993;88:791-804.
100. Ewing JA. Detecting alcoholism. The CAGE questionnaire. JAMA. 1984;252:1905-07.
101. Kessler RC, Andrews G, Mroczek Dea. The World Health Organization Composite
International Diagnostic Interview short-form (CIDI-SF). Int J Methods Psychiatr Res.
1998;7:171-85.
102. Spitzer RL, Williams JBW, Kroenke K, et al. Utility of a new procedure for diagnosing mental
disorders in primary care: The PRIME-MD 1000 study. JAMA. 1994;272:1749-56.
103. Jacobson IG, Ryan MAK, Hooper TI, et al. Alcohol use and alcohol-related problems before
and after military combat deployment. JAMA. 2008;300:663-75.
104. Brown RL, Leonard T, Saunders LA, et al. A two-item conjoint screen for alcohol and other
drug problems. J Am Board Fam Pract. 2001;14:95-106.
105. Spitzer RL, Kroenke K, Williams JB. Validation and utility of a self-report version of PRIME-
MD. JAMA. 1999;282:1737-44.
106. Proctor SP, Heaton KJ, White RF, et al. Chemical sensitivity and chronic fatigue in Gulf War
veterans: a brief report. J Occup Environ Med. 2001;43:259-64.
107. Proctor SP, Heeren T, White RF, et al. Health status of Persian Gulf War veterans: self-
reported symptoms, environmental exposures and the effect of stress. Int J Epidemiol.
1998;27:1000-10.
108. Forbes AB, McKenzie DP, Mackinnon AJ, et al. The health of Australian veterans of the 1991
Gulf War: factor analysis of self-reported symptoms. Occup Environ Med. 2004;61:1014-20.
143
109. Eisen SA, Kang HK, Murphy FM, et al. Gulf War Veterans’ Health: Medical Evaluation of a
U.S. Cohort. Ann Intern Med. 2005;142::881-90.
110. Ismail K, Kent K, Sherwood R, et al. Chronic fatigue syndrome and related disorders in UK
veterans of the Gulf War 1990–1991: results from a two-phase cohort study. Psychol Med.
2008;38:953-61.
111. Harvey SB, Hatch SL, Jones M, et al. The long-term consequences of military deployment: a
5-year cohort study of United Kingdom reservists deployed to Iraq in 2003. Am J Epidemiol.
2012;176:1177-84.
112. Harvey SB, Hatch SL, Jones M, et al. Coming home: social functioning and the mental health
of UK Reservists on return from deployment to Iraq or Afghanistan. Ann Epidemiol.
2011;21:666-72.
113. Browne T, Hull L, Horn O, et al. Explanations for the increase in mental health problems in
UK reserve forces who have served in Iraq. Br J Psychiatry Suppl. 2007;190::484-9, .
114. Milliken CS, Auchterlonie JL, Hoge CW. Longitudinal assessment of mental health problems
among active and reserve component soldiers returning from the Iraq war. JAMA. 2007
298:2141-48.
115. IOM (Institute of Medicine). Chronic multisymptom illness in Gulf War Veterans: Case
definitions reexamined. Washington, DC: Press. The National Academies Press; 2014.
116. Chalder T, Berelowitz G, Pawlikowska T, et al. Development of a fatigue scale. J Psychosom
Res. 1993;37:147-53.
117. McKenzie DP, Creamer M, Kelsall HL, et al. Temporal relationships between Gulf War
deployment and subsequent psychological disorders in Royal Australian Navy Gulf War
veterans. Soc Psychiatry Psychiatr Epidemiol. 2010;45:843-52.
118. Bolton J, Cox B, Clara I, et al. Use of alcohol and drugs to self-medicate anxiety disorders in
a nationally representative sample. J Nerv Ment Dis. 2006;194:818-25.
119. National Research Council. Gulf War Veterans: Measuring Health. Washington, DC: The
National Academies Press; 1999.
120. Wells TS, Miller SC, Adler AB, et al. Mental health impact of the Iraq and Afghanistan
conflicts: a review of US research, service provision, and programmatic responses. Int Rev
Psychiatry. 2011;23:144-52.
144
121. Chilcoat HD, Breslau N. Posttraumatic stress disorder and drug disorders: Testing causal
pathways. Arch Gen Psychiatry. 1998;55:913-17.
122. Dass-Brailsford P, Myrick AC. Psychological trauma and substance abuse: The need for an
integrated approach. Trauma Violence Abuse. 2010;11:202-13.
123. Seal KH, Cohen G, Waldrop A, et al. Substance use disorders in Iraq and Afghanistan
veterans in VA healthcare, 2001-2010: Implications for screening, diagnosis and treatment.
Drug Alcohol Depend. 2011;116:93-101.
124. Schmied EA, Highfill-McRoy RM, Crain JA, et al. Implications of psychiatric comorbidity
among combat veterans. Mil Med. 2013;178:1051-58.
125. Summerall E, McAllister TW. Comorbid posttraumatic stress disorder and traumatic brain
injury in the military population. Psychiatric Annals. 2010;40:563-80.
145